Dendritic Cell RIPK1 Maintains Immune Homeostasis by Preventing Inflammation and Autoimmunity

This article is featured in In This Issue, p.373

Receptor interacting protein kinase (RIPK) 1 is a key component of the necroptotic and apoptotic cell death pathways, and is important for the optimal activation of the NF-κB and MAPK pathways. TNF normally induces NF-κB and MAPK activation, but under certain conditions can induce apoptosis or, when caspases are inhibited, stimulate necroptosis. Necroptosis is an inflammatory form of cell death triggered by death ligands such as TNF, FasL, TRAIL, and type I and type II IFNs or by activation of pathogen recognition receptors including TLR 3 or 4 (1). RIPK1 initiates the necroptotic kinase cascade by phosphorylating and activating RIPK3, which then activates the pseudo-kinase mixed lineage kinase domain-like (MLKL) (2, 3). MLKL phosphorylation results in its translocation to the plasma membrane and changes in membrane permeability (4), resulting in the release of danger-associated molecular patterns (DAMPs) such as HMGB1, ATP, and mitochondrial DNA (5). These DAMPs activate TLRs on macrophages and dendritic cells (DCs) to induce and amplify proinflammatory cytokine and chemokine production. In some cell types, RIPK1 kinase activity is crucial for the activation of necroptosis, as the kinase inhibitor Necrostatin-1 prevents necroptosis (6) and RIPK1 kinase inactive mice, Ripk1^D138N, are resistant to necroptosis and TNF-induced shock in vivo (7, 8). RIPK1 has also been shown to have essential kinase-independent scaffold functions that mediate cell survival due to effects on the canonical (9) or noncanonical NF-κB pathways, depending on the cell type (10).

Complete RIPK1 deficiency results in postnatal lethality (9) driven by an increased sensitivity to both RIPK3-dependent necroptosis and caspase-8–dependent apoptosis, whereby compound deletion of both Caspase-8 and Ripk3 or Caspase-8 and Mlkl or the deletion of Ripk3 and TNFR type 1 (Tnfr1) rescues RIPK1-associated lethality (11–13). Whether RIPK1 triggers apoptosis and/or necroptosis is cell type and context dependent. For example, RIPK1 is essential for necroptosis in murine embryonic fibroblasts and bone marrow–derived macrophages (8, 14) but functions to negatively regulate RIPK3 and MLKL in hematopoietic stem and progenitor cells and keratinocytes (11, 15, 16). Ripk1 deletion in intestinal epithelial cells sensitizes to both TNF-mediated apoptosis and necroptosis (16). These findings reveal that RIPK1 can positively or negatively regulate necroptosis or apoptosis depending on cellular context.

DCs are critical to maintain immune homeostasis and to generate successful responses to infection. Given the important roles DCs have in maintaining tolerance, we examined the in vivo consequences of DC necroptosis on immune homeostasis by deleting Ripk1 in DCs. We found that RIPK1-deficient DCs exhibit normal responses to TNF- and FasL-induced apoptosis, but have an increased sensitivity to necroptosis in vitro. Mice with a DC RIPK1 deficiency develop inflammation and autoimmunity, characterized by splenomegaly, lymphadenopathy, tissue fibrosis, and production of anti‐nuclear autoantibodies (ANAs). We demonstrate that inflammation and autoimmunity associated with a DC RIPK1 deficiency are rescued by expression of kinase inactive RIPK1 or an absence of RIPK3, MLKL, or the TLR adapter MyD88, thereby implicating necroptosis as the mediator of inflammation, and MyD88-dependent TLR signaling as an amplifier of DAMP signaling. Importantly, autoantibody production but not inflammation was prevented by an absence of the type I IFNR or UNC93B1-dependent TLR 3, 7, and 9 signaling, thereby genetically separating signals that trigger DAMP release, inflammation, and fibrosis from those responsible for autoimmunity. This study provides genetic evidence that cell death and inflammation precede the development of autoimmunity and suggests that necroptosis kinase inhibitors may be useful in at-risk individuals to prevent autoimmune disease.

Ripk1 conditional mice (Ripk1^fl/fl) (16) were crossed with CD11cCre (Itgax-cre) (17), Ripk3^−/− mice (a gift from V. Dixit, Genentech, San Francisco), Mlkl^−/− (4), Ifngr1^−/− (18), Tnfr1^−/− (19), MyD88^−/− (20), Ifnar1^−/− (21), and Unc93b1^3d/3d (22) mice. B6.Cg-Tg(TcraTcrb)425Cbn/J (OT-II) and Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze} mice were obtained from Jackson Laboratory. All animal procedures used in this study were approved by the University of Massachusetts Medical School Institutional Animal Care and Use Committee. For antibiotic treatment, ampicillin (1 mg/ml), neomycin (1 mg/ml), ciprofloxacin (0.5 mg/ml), meropenem (0.5 mg/ml), and Grape Kool-Aid (20 mg/ml) were added to drinking water from 2 d after birth. Following weaning, ciprofloxacin was substituted with vancomycin (0.5 mg/ml). When littermate controls were not used, sex-matched control mice carrying the CD11cCre transgene were cohoused with experimental mice. For LPS-induced endotoxic shock experiments, age and sex-matched mice were i.p. injected with 5 mg/kg LPS from Escherichia coli (Sigma) and re-extracted using phenol chloroform as previously described (23).

Bone marrow–derived DCs (BMDCs) were generated by culturing bone marrow cell suspensions in 20 ng/ml recombinant GM-CSF (PeproTech) for 10 d. For necroptosis assays, BMDCs were treated with 0.1 μM SMAC mimetic (ChemieTek) and 10 μM zVAD (Enzo). For apoptosis assays, BMDCs were treated with cycloheximide (0.5 μg/ml), TNF-α (10 ng/ml), IFN-γ (10 ng/ml), or with FasL and control vesicles purified from N2-mFasL and N2-neo cell supernatant (diluted 1/40), as previously described (24). Splenic DCs were isolated from mice following treatment of the Flt3L producing melanoma line B16, using a CD11c positive selection kit (Stemcell Technologies).

To examine T cell proliferation, purified CD11c⁺ splenic DCs from mice of the indicated genotypes were incubated with OVA_323–339 or control OVA_257–264 peptide for 1 h. CD4⁺ T cells were isolated from spleens of OT-II mice using CD4 positive selection beads (Invitrogen). Isolated CD4⁺ cells were labeled with 0.5 μM CFSE (Invitrogen) and incubated with splenic DCs for 72 h. CFSE staining was examined in viable CD4⁺ cells by flow cytometry.

ANAs were detected by immunofluorescence on HEp-2 slides (Abs) as previously described (25).

Tissues were fixed in 10% formalin (Fisher Scientific). Slides were stained with H&E or Masson’s trichrome. Images were taken on an Olympus BX41 microscope using an Evolution MP 5.0 Mega-Pixel Camera (MediaCybernetics) and QCapture Pro software (QImaging).

Single-cell suspensions were stained with cell-surface Abs and DAPI (Molecular Probes) was used to distinguish between live and dead cells. Samples were run on a BD LSRII flow cytometer (BD Biosciences) and analyzed using FlowJo software (Tree Star). Cell populations were determined as follows: erythrocytes, CD71⁻ Ter119⁺; erythroblasts, CD71⁺ Ter119⁺; neutrophils, CD11b⁺ Ly-6C^int Gr-1⁺; monocytes, CD11b⁺, Ly-6C^hi and Gr-1^lo; CD4 T lymphocytes, CD4⁺ CD8⁻; CD8 T lymphocytes, CD8⁺ CD4⁻; B lymphocytes, B220⁺ CD4⁻ CD8⁻; germinal center B cells, B220⁺ Fas⁺ GL7⁺; progenitors, lineage⁻c-Kit⁺. DCs were gated according to (26).

Serum cytokines were measured using an 11-plex protein/peptide multiplex analysis (Luminex Technology) conducted by the National Mouse Metabolic Phenotyping Center at the University of Massachusetts Medical School. Chemokines and cytokines below the level of detection were assigned the value of zero. Serum Flt3L levels were measured by ELISA (R&D Systems). IFN-β ELISA kit was used as described previously (27).

RNA was prepared using TRIzol (Invitrogen) or with the RNeasy Mini kit (Qiagen). Each RNA sample was adjusted to contain the same quantity using the Nanodrop ND-1000 spectrophotometer (Thermo Scientific). RNA was used for quantitative RT-PCR using Power SYBR Green PCR Master Mix (Applied Biosystems) with the following primer pairs: β-actin, sense, 5′-TGG CAT AGA GGT CTT TAC GGA-3′, antisense, 5′-TTG AAC ATG GCA TTG TTA CCA A-3′; Iκbα, sense, 5′-TGA AGG ACG AGG AGT ACG AGC-3′, antisense, 5′-TTC GTG GAT GAT TGC CAA GTG-3′; A20, sense, 5′-CTT TCT TCA TGT CCG TGA ACA CT-3′, antisense, 5′-TTC AGG GCC TAG CTT CGA GT-3′; cIAP1, sense, 5′-TGT GGC CTG ATG TTG GAT AAC-3′, antisense, 5′-GGT GAC GAA TGT GCA AAT CTA CT-3′; cIAP2, sense, 5′-GCT GTG GCC TAA TGC TAG ACA-3′, antisense, 5′-GGA CAA TCT TGA TTT GCT CGG AA-3′; XIAP, sense, 5′-CGA GCT GGG TTT CTT TAT ACC G-3′, antisense, 5′-GCA ATT ATG GGA TAT TCT CCT GT-3′.

An anti-mouse albumin ELISA was used to measure urine protein (Bethyl Laboratories).

Cell lysates were prepared in RIPA buffer (150 mM NaCl, 50 mM Tris-HCl [pH 7.5], 1% NP40, 0.25% deoxycholate, 0.1% SDS, 1 mM EDTA), supplemented with protease inhibitors (Roche Applied Science), 1 mM DTT, 1 mM Na₃VaO₄, and 1 mM PMSF and boiled with SDS reducing sample buffer. Lysates were run on 4–12% Bis-Tris gels (Invitrogen). Membranes were probed with phospho-MLKL (Abcam), cIAP1 (Enzo Life Sciences), cIAP2 (R&D Systems), RIPK1 (BD Transduction), RIPK3 (Prosci), XIAP (MBL), β-Actin (Sigma-Aldrich), Erk 1/2 (Cell Signaling Technology), MLKL (a gift from J. Han) or phospho-IκBα (Cell Signaling Technology) Abs.

Statistical analyses were performed using GraphPad Prism software, version 6.0. Kaplan–Meier survival curves were analyzed using a log-rank test with a 95% confidence interval. A two-sided p < 0.05 was considered statistically significant for Student t tests and two-way ANOVA tests, with *p < 0.05, **p < 0.01.

To determine the effects of DC necroptosis on immune homeostasis, mice containing a conditional allele of Ripk1 (16) were crossed to CD11cCre transgenic mice (hereafter referred to as Ripk1^{DC KO}). Ripk1^{DC KO} mice were born at the expected Mendelian ratios and developed to adulthood normally. These mice exhibit a specific loss of RIPK1 expression in CD11c⁺-enriched splenic DCs by Western blot analysis (Fig. 1A).

To determine whether RIPK1 deletion altered DC survival, BMDCs were treated in vitro with necroptotic and apoptotic stimuli. Treatment with zVAD-fmk and SMAC mimetics primes cells for necroptosis by inhibiting caspase activation and inducing IAP degradation, respectively. Consistent with its negative regulatory roles in hematopoietic cells (11, 15), RIPK1-deficient DCs are more sensitive than controls to necroptotic death (Fig. 1B). In contrast, RIPK1 kinase inactive (Ripk1^D138N) DCs remained resistant to zVAD and SMAC treatment, revealing a requirement for the kinase activity of RIPK1 in TNF-mediated necroptosis in DCs.

Necroptosis is thought to be mediated by a serial phosphorylation cascade whereby RIPK1 phosphorylates RIPK3, which then phosphorylates the effector pseudo-kinase MLKL (2–4). To confirm that necroptosis was induced in RIPK1-deficient DCs, we examined DC cultures for evidence of MLKL phosphorylation. MLKL phosphorylation was detected in RIPK1-deficient DCs treated with zVAD-fmk and SMAC mimetics at early time points, indicating that a RIPK1 deficiency predisposes DCs to necroptotic death (Fig. 1C). RIPK3 overexpression has been shown to bypass the requirement for RIPK1 to drive necroptotic death (28); however, no differences in RIPK3 or MLKL expression were observed in RIPK1-deficient DCs (Fig. 1C). In fact, we demonstrate that increasing concentrations of zVAD-fmk, in the absence of SMAC mimetics or exogenous death ligands, was sufficient to induce necroptosis in RIPK1-deficient DCs (Supplemental Fig. 1A). DCs deficient for both RIPK1 and RIPK3 were not susceptible to zVAD-fmk–induced cell death, confirming that caspase inhibition alone was sufficient to trigger necroptotic death in RIPK1-deficient DCs (Supplemental Fig. 1A). In contrast, RIPK1-deficient BMDCs responded normally to apoptotic stimuli including exposure to FasL, IFN-γ, or TNF-α with cycloheximide (Supplemental Fig. 1B, 1C).

The fact that necroptosis was induced by treatment with zVAD-fmk and that SMAC mimetics were not required suggested that IAP/XIAP expression may be reduced in RIPK1-deficient DCs. Reduced cIAP1/2 and XIAP expression has been shown previously to sensitize Ag-presenting cells to necroptosis (29–31). We found cIAP2 and XIAP expression reduced in the absence of RIPK1; however, no significant changes in cIAP1 levels were observed (Supplemental Fig. 1D). Gene expression analysis revealed no differences in cIAP1/2 or XIAP mRNA levels in Ripk1^{DC KO} DCs (Supplemental Fig. 1E). Therefore, these data suggest that a RIPK1 deficiency may lead to cIAP2/XIAP protein degradation and predispose DCs to necroptosis.

Because RIPK1-deficient DCs exhibited increased susceptibility to necroptosis in vitro (Fig. 1B), decreased numbers of DCs were expected in Ripk1^{DC KO} mice. However, the loss of RIPK1 did not lead to alterations in the number of splenic conventional DCs (cDCs; CD11c⁺ MHC class II⁺) or plasmacytoid DCs (pDCs; CD11c^lo MHC class II⁺ Siglec H⁺) (Fig. 1D), even at 1 y of age (Supplemental Fig. 1F). Furthermore, no significant differences in the expression of activation markers CD80, CD86, or MHC class II were observed in cDCs or pDCs from Ripk1^{DC KO} mice compared with CD11cCre controls (Supplemental Fig. 1G, 1H). However, significant increases in bone marrow cDCs were observed (Fig. 1E) as well as increases in serum Flt3L levels (Fig. 1F), suggesting that DC progenitor activity may be stimulated in response to DC loss in vivo (32). The loss of RIPK1 had no detectable effects on BMDC production, as normal numbers of BMDC were generated from the bone marrow in vitro, indicating that RIPK1 is not required for DC proliferation and/or differentiation (Supplemental Fig. 1I).

We treated RIPK1-deficient and control DCs with TNF to examine NF-κB activation. Although not statistically significant, the IκBα phosphorylation in response to TNF was consistently less robust in RIPK1-deficient DCs than controls (Supplemental Fig. 1J). Similarly, IκBα expression was slightly dampened in TNF-treated RIPK1-deficient DCs (Supplemental Fig. 1K). We then examined the ability of RIPK1-deficient DCs to induce CD4 T cell proliferation. We found that RIPK1-deficient DCs induced proliferation of CD4 T cells, although slightly fewer T cells proliferated in response to Ag-stimulated RIPK1-deficient DCs (Fig. 1G). These data demonstrate that RIPK1-deficient DCs are able to respond to TNF and to Ag stimulation.

A caspase-8 deficiency has been shown to enhance the assembly and activation of the NLRP3 inflammasome, a signaling complex that mediates the processing of proinflammatory cytokines such as IL-1β (33). IL-1β secretion generally requires exposure to two stimuli: first, a priming agent such as LPS, which triggers pro–IL-1β synthesis; and a second signal such as ATP, which processes pro–IL-1β to its active form (34). RIPK1 deletion in macrophages, as well cIAP, XIAP, or caspase-8 depletion in DCs, has been shown to stimulate IL-1β production in response to LPS without the need for a second signal (11, 31, 35). We therefore examined IL-1β production in RIPK1-deficient DCs stimulated with LPS only or in combination with ATP and found that IL-1β processing was not significantly altered in LPS-primed RIPK1-deficient DCs in vitro (Supplemental Fig. 1L).

Given the increased sensitivity of RIPK1-deficient DCs to necroptosis in vitro, we examined the in vivo effects of RIPK1 deficiency in DCs by examining Ripk1^{DC KO} mice. Ripk1^{DC KO} mice developed splenomegaly and lymphadenopathy by 16 wk of age (Fig. 2A–E). Evidence of inflammation was observed in Ripk1^{DC KO} mice, with elevated numbers of neutrophils and inflammatory monocytes in the spleens of Ripk1^{DC KO} mice (Fig. 2C). Conversely, no differences were detected in splenic B or T lymphocyte populations (Supplemental Fig. 2B, 2C). Histological analysis revealed that the splenic architecture was disrupted in Ripk1^{DC KO} mice (Fig. 2D). Masson’s trichrome staining revealed collagen deposits in the spleen indicating fibrosis, which largely accounted for the increased spleen size. Masson’s trichrome staining also revealed evidence of fibrosis in skin and lungs of Ripk1^DC KO mice but not controls (Supplemental Fig. 2A).

In contrast to the spleen, the lymph nodes exhibited increased cellularity, largely consisting of B cells, but also with increases in DCs, neutrophils, and inflammatory monocytes (Fig. 2E). No increases in lymph node T cells were observed, nor was there evidence of T cell activation in spleen or lymph nodes (Supplemental Fig. 2B, 2C).

The bone marrow appeared largely normal (Supplemental Fig. 2D); however, Ripk1^{DC KO} mice had significantly decreased numbers of erythroblasts (Fig. 2F). Conversely, increased numbers of erythroblasts and lineage-negative, c-Kit⁺ progenitor cells were observed in the spleen (Fig. 2G, 2I) indicative of extramedullary hematopoiesis, potentially a result of the inflammation. Importantly, splenic erythrocyte numbers remained normal, indicating that the erythrocyte population did not contribute to spleen enlargement (Fig. 2H).

To further assess the inflammation in Ripk1^{DC KO} mice, we measured the levels of proinflammatory cytokines in the serum. A significant increase in serum TNFα and IFNγ was observed in Ripk1^{DC KO} mice compared with control mice (Fig. 2J, 2L). Although not significant, increases in serum IL-1β, IL-6, IL-12, KC, and MCP-1 were also observed in Ripk1^{DC KO} mice. To determine the consequences of inflammation and elevated cytokine levels, control or Ripk1^{DC KO} mice were given a low dose of LPS (5 mg/kg). Sixty percent of control mice succumbed to LPS administration with a median latency of 70 h, whereas all Ripk1^{DC KO} mice succumbed to LPS-induced endotoxic shock within 30 h (Fig. 2K). Additionally, the levels of proinflammatory cytokines TNF-α, IL-1β, and IFN-γ were significantly elevated in the sera of Ripk1^{DC KO} mice compared with controls (Fig. 2L). Thus, the proinflammatory cell death in Ripk1^{DC KO} mice sensitizes the mice to LPS-induced endotoxic shock. Interestingly, DC loss is a clinical feature of septic patients, suggesting that DC necroptosis may underlie the hyperinflammatory syndrome and immune suppression in patients with severe sepsis (36).

The lymphadenopathy and splenomegaly observed in Ripk1^{DC KO} mice suggested that these mice developed autoimmune disease. The B220⁺, CD3⁺, CD4⁻, and CD8⁻ T cells prevalent in autoimmune mouse models such as the Fas^lpr and Caspase-8^−/− Ripk3^−/− mice (37, 38) were not present in Ripk1^{DC KO} mice (Supplemental Fig. 2E). However, ANAs were detected in the serum of Ripk1^{DC KO} mice from 6 mo of age (Fig. 2M). To characterize the autoantibodies produced by these mice, we screened sera by immunofluorescent staining of HEp-2 cells. The staining intensity was similar to that observed with serum from autoimmune prone Fas^lpr mice. Abs reactive with RNA-associated autoantigens frequently exhibit a more speckled nuclear or cytoplasmic staining pattern, whereas Abs reactive with dsDNA or other chromatin components exhibit a homogenous nuclear stain as observed with sera from Fas-deficient mice. The staining pattern of the circulating ANAs in Ripk1^{DC KO} mice was consistent with Abs targeting nucleolar components, with one of six mice examined exhibiting a cytoplasmic staining pattern (Fig. 2M, 2N). Antinucleolar and other speckled nuclear patterns are commonly found in TLR9-deficient or TLR7-overexpressing lupus-prone mice (39, 40). These data indicate that autoantibodies produced by Ripk1^{DC KO} mice recognize RNA-associated autoantigens and not dsDNA.

Correlating with the presence of ANAs, we observed increased numbers of germinal center B cells in the lymph nodes of these mice (Supplemental Fig. 2F). Although ANAs were detected in Ripk1^{DC KO} mice, the CD11c promoter can be active in some lymphoid cells (17), therefore we mated the Ripk1^{DC KO} and CD11cCre mice with a Rosa26-loxP-Stop-loxP-tdTomato reporter line and assessed Tomato red expression in hematopoietic cells from spleen and lymph nodes (Supplemental Fig. 2G). Tomato red expression was detected in 86% of DCs and 26% of B cells isolated from spleen and lymph nodes of Ripk1^{DC KO} but not control mice. To assess Ripk1 deletion in B cells, we sorted Tomato red–positive and –negative B cells from control and Ripk1^{DC KO} mice and performed genomic PCR. We detected evidence of Ripk1 deletion in Tomato red–positive but not –negative B cells (Supplemental Fig. 2I). Notably, B cells isolated from mice with lymphadenopathy retained a floxed Ripk1 allele, indicating that these B cells likely expressed RIPK1. Based on this analysis and the fact that germinal center B cells expanded in the lymph node, we conclude that Ripk1 deletion in the CD11c-positive, age-associated B cell population associated with young lupus-prone or aged wild-type mice is not responsible for B cell expansion and autoantibody production in these mice (41). As is typical for mice on the C57BL/6 background (42), Ripk1^DC KO mice did not develop proteinuria by 12 mo of age (Supplemental Fig. 2J). Together, these data suggest that a RIPK1 deficiency in DCs leads to systemic inflammation that triggers B cell expansion and autoantibody production.

Given the in vitro sensitivity of RIPK1-deficient DCs to necroptosis (Fig. 1B) and the lack of reliable and sensitive measures to detect necroptotic death in mice, we used a genetic approach to determine whether aberrant DC necroptosis is responsible for the inflammation and autoimmunity in Ripk1^{DC KO} mice. To address the specific role of RIPK1 kinase-dependent cell death, Ripk1^{DC KO} mice were crossed to mice expressing the kinase inactive RIPK1 (D138N) allele (8) to generate CD11cCre Ripk1^fl/D138N (Ripk1^{DC KO/D138N}) mice. Because the kinase activity of RIPK1 is not required for necroptosis in certain cell types (43, 44), we also generated Ripk1^{DC KO} Ripk3^−/− and Ripk1^{DC KO} Mlkl^−/− mice to provide additional genetic evidence that DC necroptosis results in chronic inflammation and a break in tolerance.

Introduction of the Ripk1^D138N allele or deletion of Ripk3 protected RIPK1-deficient DCs from necroptosis in vitro (Fig. 3A). Similarly, the introduction of the Ripk1^D138N allele or deletion of Ripk3 or Mlkl ameliorated the inflammatory and autoimmune disease associated with a DC RIPK1 deficiency, resulting in reductions in spleen and lymph node weight (Fig. 3B). Significant decreases in neutrophil and inflammatory monocyte infiltration were also observed in the spleens of these mice indicating that inflammation was reduced (Fig. 3C). Importantly, the splenic architecture was maintained in these mice with no detectable signs of fibrosis (Fig. 3D, Supplemental Fig. 3A). Interestingly, some differences were observed in Ripk1^{DC KO/D138N}, Ripk1^{DC KO} Ripk3^−/−, and Ripk1^{DC KO} Mlkl^−/− mice. For example, although Ripk1^{DC KO/D138N} and Ripk1^{DC KO} Mlkl^−/− mice showed a complete rescue of the lymphadenopathy, Ripk1^{DC KO} Ripk3^−/− mice displayed only a partial reduction in lymph node weight and cellularity, with increased numbers of inflammatory monocytes, neutrophils, and B cells in the lymph nodes (Fig. 3E). Collectively, these data demonstrate that chronic necroptosis drives inflammation in Ripk1^{DC KO} mice. However, these data are also consistent with published work that reveals necroptosis-independent functions of RIPK3 (33, 35, 45).

We predicted that the release of DAMPs from necroptotic DCs drive the autoimmunity observed in Ripk1^{DC KO} mice. Therefore, we examined autoantibody production in Ripk1^{DC KO/D138N}, Ripk1^{DC KO} Ripk3^−/−, and Ripk1^{DC KO} Mlkl^−/− mice. We were unable to detect ANAs in the serum of these mice (Fig. 3F, Supplemental Fig. 3B), suggesting that necroptotic DCs release DAMPs including RNA and other nucleolar components that trigger B cell proliferation, differentiation, and autoantibody production. The ability of a RIPK1 kinase inactive D138N allele, an MLKL deficiency and to some extent a RIPK3 deficiency, to rescue disease in Ripk1^{DC KO} mice provides genetic evidence that DC necroptosis underlies the inflammation and autoimmunity.

TNF-α and IFN-γ are known to induce necroptosis (1) and elevated levels of TNF-α and IFN-γ were observed in the serum of Ripk1^{DC KO} mice (Fig. 2J), implicating TNF-α and/or IFN-γ as inducers of DC necroptosis in Ripk1^{DC KO} mice. The increased sensitivity of RIPK1-deficient BMDCs to SMAC- and zVAD-induced necroptosis in vitro was lost upon compound deletion of Tnfr1, but not Ifngr1 (Fig. 4A), further implicating TNF-α as the DC necroptosis-inducing ligand in Ripk1^{DC KO} mice.

To address the roles of TNF-α and IFN-γ in inflammation and autoimmunity, we generated Ripk1^{DC KO} mice that lack either TNF-R1 or IFN-γR1. Surprisingly, the absence of TNF-α or IFN-γ signaling had little effect on spleen inflammation, with both Ripk1^{DC KO} Tnfr1^−/− and Ripk1^{DC KO} Ifngr1^−/− mice exhibiting splenic fibrosis and significant increases in inflammatory cell infiltration (Fig. 4B–D). Consistent with the accelerated TNF-mediated necroptosis observed in vitro (Fig. 4A), we found the lymphadenopathy significantly ameliorated in Ripk1^{DC KO} Tnfr1^−/− mice but not in Ripk1^{DC KO} Ifngr1^−/− mice (Fig. 4B, 4C). The decrease in lymph node size observed in Ripk1^{DC KO} Tnfr1^−/− mice appeared due to decreases in B cell number (Fig. 4E; p = 0.075), indicating that TNF-R1 signaling is required for optimal germinal center B cell expansion in lymph nodes. Serum ANAs, however, were not detected in either mouse strain, indicating that loss of TNF-α or IFN-γ signaling is sufficient to prevent ANA development (Fig. 4F, Supplemental Fig. 3C).

Necroptotic death is predicted to result in the release of cytokines, chemokines, and DAMPs that activate TLRs and NOD-like receptors to promote inflammation (5). We examined the contribution of MyD88, a key component of TLR signaling (46), to the inflammatory disease observed in Ripk1^{DC KO} mice. At 8 wk of age Ripk1^{DC KO} Myd88^−/− mice exhibited significant decreases in spleen and lymph node size compared with Ripk1^{DC KO} mice, together with decreases in neutrophil and inflammatory monocyte infiltrates in the spleen and lymph nodes (Fig. 5A–C). In addition, the number of B cells in the lymph nodes was comparable to controls. These data strongly suggest that the inflammation observed in Ripk1^{DC KO} mice is exacerbated by MyD88-dependent TLR signaling, which is likely acting to amplify inflammation in response to the release of DAMPs from necroptotic RIPK1-deficient DCs. Notably, a TLR-MyD88 pathway also drives the B cell expansion observed in the lymph nodes of Ripk1^{DC KO} mice.

The MyD88 dependency raised the possibility that pathogenic or commensal bacteria trigger and/or exacerbate inflammation in mice with DC RIPK1 deficiency. To examine this, we treated Ripk1^{DC KO} mice with broad-spectrum antibiotics for 12 wk. We found that antibiotic treatment had no detectable effect on the development of inflammatory disease in Ripk1^{DC KO} mice (Fig. 5D–F). This suggests that the MyD88-dependent inflammation in Ripk1^{DC KO} mice is independent of pathogenic or commensal bacteria and likely a consequence of endogenous DAMPs.

The Ripk1^{DC KO} Myd88^−/− mice began to lose weight and became moribund, succumbing with an average latency of 11.5 wk (Fig. 5G). Treatment with broad-spectrum antibiotics significantly prolonged the survival of Ripk1^{DC KO} Myd88^−/− mice (Fig. 5G). The heightened susceptibility of Ripk1^{DC KO} Myd88^−/− mice to bacterial infection suggests that a combined total MyD88 deficiency with a DC RIPK1 deficiency interfered with the animals’ innate and adaptive antibacterial immune responses.

Given the recent data implicating type I IFNs in necroptosis and known roles for type I IFNs in ANA production and autoimmunity (47, 48), we hypothesized that type I IFN may contribute to autoinflammation and autoimmunity observed in Ripk1^{DC KO} mice. Although we were unable to detect IFN-α or IFN-β in the serum of Ripk1^{DC KO} mice (Supplemental Table I), we tested this hypothesis genetically by generating Ripk1^{DC KO} Ifnar1^−/− mice. The absence of the type I IFNR had no effect on the splenomegaly or lymphadenopathy associated with a DC RIPK1 deficiency, ruling out a critical role for type I IFN-induced necroptosis in the organ inflammation (Fig. 6). Although inflammation remained unaffected, an absence of type I IFN signaling prevented ANA development in five of six Ripk1^{DC KO} Ifnar1^−/− mice examined (Fig. 6E, Supplemental Fig. 3D). One Ripk1^{DC KO} Ifnar1^−/− mouse exhibited weak ANA staining (Supplemental Fig. 3D). Surprisingly, B cell numbers remained increased in the lymph nodes of Ripk1^{DC KO} Ifnar1^−/− mice. Overall, these data reveal that autoimmunity but not inflammation is largely driven by or dependent on type I IFN signaling. These data provide genetic evidence that inflammation precedes autoimmunity, and suggests that type I IFNs mediate B cell differentiation into Ab-secreting plasma cells.

Given that the ANAs generated in Ripk1^{DC KO} mice largely target the nucleolar components of the cell (Fig. 2N), we examined the role of the nucleic acid sensing TLRs as drivers of the autoimmune disease. We used the Unc93b1^3d/3d mouse, which lacks functional UNC93B1 required for the intracellular trafficking of TLRs 3, 7, and 9 from the endoplasmic reticulum to the endosome (22). These TLRs sense nucleolar components such as ssRNA, dsRNA, and unmethylated DNA and TLR 7 is required for the development of autoimmunity in lupus-prone mouse strains (22, 49, 50). An UNC93B1 deficiency had no detectable effects on splenomegaly, Masson’s trichrome staining, or on splenic inflammatory cell infiltrates, indicating that the nucleic acid TLRs do not mediate the inflammation and tissue fibrosis (Fig. 7). The size and cellularity of the lymph nodes, however, were significantly reduced in Ripk1^{DC KO} Unc93b1^3d/3d mice (Fig. 7A). Importantly, the UNC93B1-dependent reduction in lymph node size correlated with significant reductions in B cell number, whereas neutrophils and inflammatory monocytes were not significantly altered (Fig. 7D). No ANAs were detected in serum from Ripk1^{DC KO} Unc93b1^3d/3d mice (Fig. 7E, Supplemental Fig. 3E), revealing that nucleic acid sensing TLRs are required for the generation of autoantibodies.

These genetic data reveal that the inflammation, tissue fibrosis, and autoimmunity is dependent on the presence of RIPK1 kinase activity, RIPK3, MLKL, and MyD88 (Figs. 3, 5, 6). Necroptotic cell death and MyD88-dependent TLRs mediate the inflammatory disease; however, the nucleic acid sensing TLRs do not contribute to tissue fibrosis or inflammation. Although type I and II IFNs induce necroptosis in vitro and are thought to underlie TNF-induced necroptosis in vivo (12, 51), neither IFN-γ or IFN-α/β are major drivers of inflammatory disease in this model. By contrast, the B cell expansion is partially TNF-R1–dependent, and requires both MyD88 and UNC93B1. Surprisingly, type I IFN signaling is not required for B cell expansion in the lymph node, but proves essential for the differentiation of these cells into Ab-secreting cells and for ANA production. These in vivo genetic studies delineate the cell death and innate immune pathways as well as the temporal order of events that break tolerance and give rise to autoimmunity. These findings reveal that DC necroptosis is sufficient to give rise to inflammation and autoimmunity.

Herein we demonstrate that the loss of RIPK1 primes DCs for RIPK3- and MLKL-dependent necroptosis leading to inflammation, tissue fibrosis, and autoimmunity. We show that in DCs, RIPK1 negatively regulates necroptosis but has no detectable effects on apoptosis. Genetically blocking necroptosis in DCs prevents the development of inflammation and autoimmunity in Ripk1^{DC KO} mice, revealing a crucial role for RIPK1 in DC survival and maintenance of immune homeostasis.

The proinflammatory cytokine TNF is considered the classical activator of necroptosis. We show that in vitro RIPK1-deficient DCs undergo TNF-mediated necroptosis (Figs. 1B, 4A); however, an absence of TNF-R1 had no detectable effects on splenic inflammation or tissue fibrosis (Fig. 4). These findings were unexpected as pulmonary fibrosis patients display elevated levels of TNF-α and mice that overexpress TNF in the lung develop progressive pulmonary fibrosis (52). Splenic fibrosis was prevented by expression of kinase inactive RIPK1 or an absence of RIPK3 or MLKL (Fig. 3D), implicating necroptosis in the fibrotic response. An absence of TNF-R1 signaling significantly ameliorated the lymphadenopathy in Ripk1^{DC KO} mice (Fig. 4B), potentially due to TNF-R1 effects on follicular DC networks and/or germinal center formation (53, 54).

Because an absence of TNF, IFN-γ, or type I IFN signaling had no effect on inflammation, we speculate that ligand-independent DC necroptosis drives the inflammation observed in Ripk1^{DC KO} mice. A RIPK1 deficiency in DCs stimulates RIPK3-MLKL–dependent necroptosis in vivo; however, no detectable increases in RIPK3 and/or MLKL expression were observed in vitro (Fig. 1C). The necroptotic sensitivity of RIPK1-deficient DCs could be explained by decreased expression of cIAP2 and XIAP (Supplemental Fig. 1D). Consistent with this hypothesis, cIAP1 and cIAP2 depletion in macrophages and monocytes have been show to stimulate necroptosis (29, 30). The IAP proteins regulate RIPK1 polyubiquitination and reductions in polyubiquitinated RIPK1 stimulate the formation of RIPK1-containing death complexes in the cytosol. Thus, a RIPK1 deficiency, which may result in IAP degradation (10, 16, 55), may sensitize DCs to necroptosis by unleashing RIPK3 and by promoting necrosome or complex 2b (RIPK3/MLKL) formation. XIAP loss in DCs also alters RIPK1 polyubiquitination and promotes RIPK3-mediated cell death by as yet unclear mechanisms (31).

The deletion of other death receptor pathway components in DCs such as caspase-8 or Fas-associated death domain protein (termed FADD) reveal similar but distinct disease phenotypes. For example, although the deletion of FADD sensitized the DCs to necroptosis leading to inflammation in vivo, the inflammation was driven by commensal bacteria (56). Yet, depleting commensal bacteria in Ripk1^{DC KO} mice had no effect on disease severity or kinetics (Fig. 5D–F). A caspase-8 DC deficiency also results in chronic inflammation that was rescued upon deletion of RIPK3. However, the effects on inflammatory disease were not attributed to necroptosis, but rather to requirements for caspase-8 and RIPK3 in inflammasome activation and IL-1β production (33, 57). In contrast to caspase-8–deficient DCs, we found RIPK1-deficient DCs are not primed to produce higher levels of IL-1β when stimulated with LPS in vitro (Supplemental Fig. 1L). Collectively these data suggest that commensal bacteria or inflammasome activation are not responsible for disease in Ripk1^{DC KO} mice.

Mice deficient for caspase-8 are protected from embryonic lethality by concomitant loss of RIPK3 or MLKL, revealing that caspase-8 prevents necroptosis during development (38, 58, 59). The Caspase-8^−/− Ripk3^−/− and Caspase-8^−/− Mlkl^−/− mice are viable and fertile but develop autoimmune disease due to the expansion of CD3⁺, B220⁺, CD4⁻, and CD8⁻ T cells, due to impaired Fas-mediated apoptosis. We have shown that RIPK1 functions in certain cell types to prevent RIPK3 and MLKL activation; however, this unusual T cell population was not detected in autoimmune Ripk1^{DC KO} mice nor did we find any evidence of T cell activation (Supplemental Fig. 2).

Ripk1^{DC KO} mice develop autoimmunity, which is prevented when necroptosis is blocked. We show that autoimmunity is type I IFN dependent and that inhibiting TLR3, 7, and/or 9 signaling largely prevents ANA production. Interestingly, in our model, blocking the nucleic acid sensing TLRs (likely TLR7/9), but not type I IFNs, inhibits B cell expansion in the lymph nodes. The fact that B cell numbers were not altered in Ifnar1^−/− mice was unexpected, as TLR7/9 play central roles in autoimmune disease and one of the main downstream effects of TLR7/9 signaling is IFN-α/β production (60). B cells may directly respond to RNA or DNA/immune complexes through the BCR which can directly bind ssRNA or dsDNA and transport it to the endosome to activate TLR7 or TLR9, respectively (61, 62). Collectively, these data indicate that endosomal TLR7 and/or TLR9 cooperate with the BCR to mediate the expansion of autoreactive B cells via a pathway that does not require IFN-α/β.

The frequency of Ripk1^{DC KO} mice that produce ANAs targeting nucleolar components correlate with data showing that anti-RNA autoantibodies are common in systemic lupus erythematosus (SLE) patients (63, 64). Our studies suggest that ANA production in SLE patients may also be a consequence of increased necroptosis, which likely provides a continuous source of RNA autoantigens. Finally, these data demonstrate that chronic necroptosis (in the absence of DC or T cell activation) is sufficient for germinal center B cell expansion and ANA production.

A theme that has emerged from the study of a wide range of autoimmune diseases is the appearance of disease-specific autoantibodies before evidence of clinical disease. In rheumatoid arthritis, Abs reactive with citrullinated histones can be detected years before any evidence of joint inflammation and IFN-γ levels correlate with the appearance of serum autoantibodies years prior to the clinical diagnosis of SLE (65–68). These genetic studies reveal that necroptotic cell death triggers inflammation and precedes the development of autoimmunity. Tissue fibrosis observed in the spleen, lungs, and skin of Ripk1^{DC KO} mice was prevented by expression of kinase inactive RIPK1 or an absence of RIPK3 or MLKL, implicating, to our knowledge, for the first time, this form of inflammatory cell death in the fibrotic response. Fibrosis was unaffected by an absence of TNF-R1, type I or II IFN signaling, but appeared rescued by an MyD88 deficiency. Our data also show that necroptosis and inflammation are not sufficient for the development of autoimmunity; ANA production requires the presence of functional nucleic acid–sensing TLRs and the type I IFNR.

Recently, the inflammation associated with a RIPK1 deficiency in keratinocytes has been shown to be dependent on the receptor interacting protein homotypic interacting motif of RIPK1, MLKL, and the putative nucleic acid sensor Z-DNA binding protein 1, ZBP1, also known as DNA-dependent activator of IFN regulatory factors (69, 70). These findings suggest that the ZBP1-RIPK3-MLKL necroptotic pathway may be constitutively activated in Ripk1^{DC KO} mice. Although initially identified as a dsDNA sensor capable of activating the NF-κB and IRF3 transcription factors (71), ZBP1 has recently been shown to recognize RNA viruses and induce necroptosis (72). Interestingly, ZBP1 expression is increased in PBMCs of lupus patients (73) raising the intriguing possibility that ZBP1-mediated necroptosis may contribute to SLE and other autoimmune or fibrotic disorders.

We thank Dr. Ben Croker, Dr. Shruti Sharma, and Dr. Kristen Nundel for invaluable advice and discussion. We thank Dr. Dave Garlick for histopathological analysis. We thank Jason McGowan and Stacy Cote for help with ANA experiments. We acknowledge the support of the University of Massachusetts Flow Cytometry Core and the Mouse Metabolic Phenotyping Center.

The authors have no financial conflicts of interest.