Glutathione peroxidase 4 regulated neutrophil ferroptosis induces systemic autoimmunity

Serum factors modulate neutrophil viability in SLE

To substantiate the concept that neutropenia is a common feature in SLE, a total of 126 patients with SLE were included in the analysis (Supplementary Table 1-3). Compared with healthy controls (HC), neutrophil counts in patients with SLE were significantly reduced, with 35% of them presenting a neutrophil count below 2 × 10⁹/L. In contrast, this was not observed in patients with other autoimmune diseases (Fig. 1a). Moreover, to exclude the potential interference of treatment, 98 patients with newly diagnosed untreated SLE were selected and as expected, the neutrophil counts correlated inversely with disease activity as measured by SLE Disease Activity Index (SLEDAI) (Fig. 1b). Importantly, the numbers of neutrophils were restored to normal levels after effective treatment with standard of care medications. (Fig. 1c, Supplementary Table 4).

We next evaluated the viability (live cells were gated as 7AAD⁻ Annexin V⁻) of fresh neutrophils isolated from SLE patient blood and as expected, they displayed lower cell viability compared with those from HCs; notably, effective treatments helped restore neutrophil viability (Fig. 1d). Collectively, these observations led us to hypothesize that the sera from patients with active SLE might promote neutrophil death. As expected, neutrophils cultured with SLE sera displayed significantly reduced viability (Fig. 1e, Extended Data Fig. 1a) and this reduction correlated positively with the severity of neutropenia in the patients (Fig. 1f and Supplementary Table 5). Of note, serum from RA patients did not have the same effect (Extended Data Fig. 1b). To determine key factors in SLE sera responsible for this effect, we compared cytokine profiles in SLE sera with those in other autoimmune diseases and HCs. Four cytokines were specifically increased in SLE, including IFN-α, CXCL11, IL-12p40 and IL-23 (Fig. 1g,h, Extended Data Fig. 1c). However, blockade of type 1 IFN signaling but not of the other cytokines¹², abrogated the enhanced neutrophil death mediated by SLE sera (Fig. 1i, Extended Data Fig. 1d), and the addition of IFN-α to HC serum promoted neutrophil death (Fig. 1j). These results indicate that type 1 IFN contributes to neutrophil death in lupus patients.

The production of autoantibodies is a hallmark of SLE and plays a critical role in disease pathogenesis^{13, 14}. Interestingly, the proportion of autoantibodies, such as anti-dsDNA IgG, in total IgG correlated positively with the severity of neutropenia (Extended Data Fig. 1f,g). To investigate whether autoantibodies are involved in neutrophil death, we purified total IgG from either healthy or SLE sera (Extended Data Fig. 1h,j), and applied them to healthy neutrophils in the presence of HC serum. As expected, SLE but not healthy IgG reduced the viability of neutrophils in a concentration-dependent manner (Fig. 1k, Extended Data Fig. 1k). Consistently, the depletion of IgG from SLE serum reduced the ability of serum to decrease cell viability of cultured neutrophils and the addition of IFNAR neutralizing antibody further curtailed this modulating capability (Fig. 1l, Extended Data Fig. 1i,k). Taken together, these results suggest that SLE IgG and IFN-α present in SLE sera induce neutrophil death.

Neutrophil ferroptosis is prevalent in patients with SLE

NETosis, a unique form of neutrophil death characterized by the release of decondensed chromatin and granular contents to the extracellular spaces, has been proposed as an important cause of neutropenia in lupus¹⁵. To test this, peripheral neutrophils from both HC and patients with SLE were stained with propidium iodide (PI) to detect dead cells and DAPI as counterstain, and NETs were defined as cells with a surrounding DNA area exceeding 400 μm^{2 16}. We observed only a small portion of dead neutrophils could be attributed to NETosis (Fig. 2a). Therefore, we further looked into neutrophil death by electron microscopy, and we found a significant portion of neutrophils isolated from patients with SLE exhibited typical morphological characteristics of ferroptosis⁵, including mitochondrial vacuole formation with increased mitochondrial membrane density and disappearance of mitochondrial cristae (Fig. 2b). Similar morphologic changes were recapitulated in HC neutrophils treated with RSL-3, a ferroptosis inducer¹⁷ (Fig. 2c). Consistent with this, the levels of lipid-ROS, both an indicator and strong inducer of ferroptosis⁵, were higher in patients with SLE with active diseases but were remarkedly reduced after effective systemic treatments (Fig. 2d). Of note, the increased levels of lipid-ROS were restricted to neutrophils and did not extend to lymphocytes or monocytes from patients with SLE, and consistently SLE serum induced lipid-ROS and reduced cell viability in neutrophils only (Fig. 2e,f, Extended Data Fig. 2a,b). In contrast to the finding that RSL-3 induced death in neutrophils cultured with HC serum, two ferroptosis inhibitors, liproxstatin-1 (LPX-1) and the iron chelator deferoxamine (DFO), rescued neutrophil death induced by SLE serum ^{5, 18} (Fig. 3a,b, Extended Data Fig. 2c-e), suggesting that ferroptosis may be the main form of neutrophil death in SLE. Of note, LPX-1 did not inhibit NETosis and had no or negligible effects on B cell or plasmacytoid dendritic cell function indicated by antibody or type I IFN production, respectively (Fig. 3c, Extended Data Fig. 2f,g 3a-e). Notably, necroptosis inhibitors, necrosulfonamide (NSA) and necrostatin-1 (Nec-1), and the apoptosis inhibitor, Z-VAD, produced only minimal effects on neutrophil death (Fig. 3b). It has been shown that direct inhibition of GSH synthesis induces ferroptosis¹⁹. Of note, low dose of ß-ME could partially rescue neutrophil death by boosting GSH biosynthesis (Extended Data Fig. 2h,i). Collectively, these findings indicate that ferroptosis is the main driver of neutrophil death in patients with SLE.

As was expected, the addition of SLE IgG or IFN-α to HC serum increased neutrophil lipid-ROS in a concentration-dependent manner (Fig. 3d,e), whereas IgG depletion or treatment with IFNAR neutralizing antibodies reduced the capacity of SLE sera to induce lipid-ROS production by neutrophils (Fig. 3d,f and Extended Data Fig. 1e). Although both IFNα and SLE IgG induced neutrophil ferroptosis, SLE IgG was a more potent inducer. Of note, although there was approximately 20% NETosis in total dead neutrophils induced by IFN-α stimulation alone, the addition of SLE IgG, clearly re-programed the neutrophil death program indicated by a significant decrease of NETosis in the face of slightly enhanced neutrophil death (Extended Data Fig. 3f-i). Taken together, our results suggest that in the SLE milieu, neutrophil ferroptosis is a dominant form of neutrophil death considering the co-presence of IFNα and IgG.

Inhibition of ferroptosis attenuates lupus in mice

We next examined neutrophil ferroptosis in different lupus-prone mice^{20, 21}. Consistent with our finding in patients with SLE, we observed decreased neutrophil viability and increased lipid-ROS production in both MRL/lpr and NZB/W F1 mice (Fig. 4a,b). Furthermore, LPX-1 treatment of MRL/lpr mice at 12 weeks of age efficiently inhibited the production of lipid-ROS in neutrophils, significantly mitigated disease progression, reduced the production of autoantibodies and various inflammatory cytokines, as well as increased serum complement component 3 (C3), alleviated splenomegaly, lymphadenopathy and severity of lupus nephritis (Fig. 4c-j, Extended Data Fig. 4a-i). Cytoxan (CTX), a drug used to treat people with SLE was administered as control²². Of note, the NETosis inhibitor Cl-amidine did not provide equivalent therapeutic values as LPX-1 confirming the distinct role of neutrophil ferroptosis in lupus pathogenesis (Extended Data Fig. 4j-o). Taken together, these results suggest that neutrophil ferroptosis is a main cause of neutropenia in lupus and targeted therapies and correction of this abnormality yield therapeutic effects.

SLE IgG and IFN-α suppress GPX4 expression in neutrophils

We next conducted RNA-Seq analysis and compared neutrophil ferroptosis related gene expression between HCs and patients with SLE, and identified 21 down-regulated genes and 2 upregulated genes in SLE neutrophils (Extended Data Fig. 5a, Supplementary Table 6,7). Given that GPX4 has been reported to be a key negative regulator of ferroptotic cell death by removing lipid peroxides ^{6, 23, 24}, we confirmed that GPX4 expression was significantly decreased in the neutrophils but not other immune cells from both lupus-prone mice and people with SLE (Fig. 5a-e, Extended Data Fig. 6a-e, Supplementary Table 5) although the expression of SLC7A11 was not significantly changed (Extended Data Fig. 5b,c). The expression of GPX4 in neutrophils correlated inversely with SLEDAI score but was restored to normal levels after effective systemic treatments (Fig. 5d, Extended Data Fig. 6b, Supplementary Table 5). Furthermore, when we examined GPX4 expression in neutrophils cultured with different sera, we found, not surprisingly, GPX4 expression was downregulated when cultured with SLE serum, especially from patients with active disease and neutropenia, but not HC serum (Fig. 5f-h, Extended Data Fig. 6h,i).

As autoantibodies and type I IFNs were found to be pivotal in inducing neutrophil ferroptosis, we next examined their effect on GPX4 expression in neutrophils. As expected, we observed that autoantibodies and IFN-α suppressed the expression of GPX4, whereas addition of anti-IFNAR or IgG depletion reduced the ability of SLE serum to downregulate GPX4 (Fig. 5i-l). Of note, in healthy donors, neutrophils express significantly lower GPX4 compared to other immune cells, which may explain why neutrophils are more sensitive to SLE serum induced ferroptosis considering their much lower threshold of ferroptosis induction (Extended Data Fig. 6f,g). Of note, GPX4 expression was not affected by NETosis inhibitors (Extended Data Fig. 6j,k). The expression correlation analysis indicated the involvement of FcγR3β in neutrophils ferroptosis by regulating GPX4 (Extended Data Fig. 7a,b). The high expression of FcγR3β in neutrophils aligned with the finding that the overexpression of FcγR3β significantly enhanced the sensitivity of HL60 cells to SLE IgG induced GPX4 reduction confirmed the role of FcγR3β in SLE IgG mediated neutrophil ferroptosis (Extended Data Fig. 7c-e). Because TLR-dependent mechanisms have been shown to trigger NETosis ^{3, 25}, more information is needed to exclude their relevant contribution to neutrophil ferroptosis. Together, our results demonstrate that the autoantibodies and the increased IFN-α in SLE sera downregulate the expression of GPX4 and lead to neutrophil ferroptosis.

Gpx4^fl/wt LysMCre⁺ mice develop lupus-like disease

To validate the pathogenic role of neutrophil ferroptosis in vivo, we generated a myeloid cell-specific Gpx4 haploinsufficient mouse (Gpx4^fl/wt LysMCre⁺) to examine the direct contribution of defective GPX4 expression in neutrophils to the breakdown of immune tolerance. As expected, Gpx4^fl/wt LysMCre⁺ mice showed reduction of GPX4 expression in neutrophils (Extended Data Fig. 8a,b). Of note, the percentage and absolute cell count of neutrophils in the peripheral blood were significantly decreased in Gpx4^fl/wt LysMCre⁺ mice, which mimicked the neutropenia in patients with SLE (Extended Data Fig. 8c,d). The increased lipid-ROS production and decreased neutrophils viability in these mice phenocopied what we noted in patients with SLE, and also these defects could be rescued by addition of LPX-1 (Extended Data Fig. 8e). Furthermore, we observed lupus-like manifestations, including the production of inflammatory cytokines and autoantibodies, development of skin lesion, splenomegaly, lymphadenopathy, proteinuria and glomerular deposition of IgG, IgM and C3 in Gpx4^fl/wt LysMCre⁺ mice (Fig. 6a-e, and Extended Data Fig. 8f,g), Interestingly, Gpx4^fl/fl LysMCre⁺ mice with homozygous GPX4 reduction exhibited only mild signs of autoimmunity with extremely low neutrophils in vivo, which was consistent with the fact that neutrophils are requisite in lupus pathogenesis. (Extended Data Fig. 8h-j). Collectively, the Gpx4^fl/wt LysmCre⁺ mice represent a new mouse to study lupus and the findings herein strongly support the role of neutrophil ferroptosis in the immunopathogenesis in SLE.

SLE IgG and IFN-α promote CREMα binding to the Gpx4 promoter

Next, we aimed to explore the molecular pathway regulating ferroptosis in neutrophils. By nucleotide sequence analysis, we found a conserved cAMP response element (CRE) located at 42~35 base pairs upstream of the human Gpx4 promoter, which was a critical binding site for cAMP-response element-binding protein (CREB) and cAMP response element modulator (CREM)^{26, 27}. CREMα is a widely expressed transcriptional repressor that has been implicated in the termination of T cell immune responses, and increased CREMα in SLE T cells has been linked to decreased interleukin-2 (IL-2) and increased interleukin-17F (IL-17F) production^{28, 29}. Calcium/calmodulin kinase IV (CaMKIV) has been demonstrated to be involved in the translocation of CREMα to the nucleus and its binding to the CRE sites³⁰. As expected, we observed increased nuclear accumulation of both CaMKIV and CREMα in SLE neutrophils (Fig. 7a, Extended Data Fig. 9a). This phenomenon could also be induced in HC neutrophils by culture with SLE serum, or with HC serum supplemented with either SLE IgG or IFN-α (Fig. 7b,c). In contrast, SLE serum with either depletion of IgG or addition of IFNAR neutralizing antibodies failed to induce nuclear translocation of these two molecules (Fig. 7b,c). These data indicate that autoantibodies and type 1 IFNs induce nuclear translocation of both CaMKIV and CREMα in SLE neutrophils.

In agreement with the increased CREMα nuclear accumulation, the enhanced binding of CREMα to the site 42 base pair upstream of Gpx4 promoter was observed in SLE neutrophils (Fig. 7d). In addition, enrichment of CREMα correlated positively with the activation of type I IFN signaling (Fig. 7e). Moreover, SLE neutrophils displayed increased binding of nuclear CREMα to the site 42 base pair upstream of the Gpx4 promoter as assessed by DNA pulldown assay (Fig. 7f). Accordingly, the binding of CREMα to the Gpx4 promoter was increased in HC neutrophils cultured with SLE serum, or HC serum supplemented with either SLE IgG or IFN-α, whereas the binding complex was reduced in SLE serum mixed with IFNAR neutralizing antibodies or depleted of IgG (Extended Data Fig. 9b).

Finally, siRNA knockdown of CREMα in neutrophil-like HL60 cells significantly abrogated GPX4 inhibition by SLE serum, autoantibodies, or IFN-α; in contrast, overexpression of CREMα decreased GPX4 expression (Extended Data Fig. 9c-e), supporting that CREMα is a key regulator upstream of GPX4 in neutrophils. As expected, HL60 cells stimulated with these factors displayed consistent changes in lipid-ROS (Extended Data Fig. 9f).

To validate the CREMα/ CaMKIV axis in neutrophil ferroptosis in vivo, we examined neutrophils in Camk4^−/−MRL/lpr mice and observed higher levels of GPX4 and cell viability and lower lipid-ROS (Fig. 7g-i) compared with matched wild type controls, confirming that CaMKIV deficiency promoted GPX4 expression and reduced neutrophil ferroptosis. Furthermore, we induced lupus-like disease in wild type, CREM-deficient (Crem^−/−) and CaMKIV-deficient (Camk4^−/−) mice on the C57BL/6 background by administering pristane and adenovirus IFNα^{31, 32}. As expected, the combined effect of autoantibodies and IFNα significantly decreased neutrophil GPX4 expression and viability and increased the production of lipid-ROS in wild type but not Crem^−/− and Camk4^−/− mice (Fig. 7j-m). Together, our findings demonstrate that autoantibodies and IFN-α in SLE promote the nuclear translocation of CREMα and its binding to the Gpx4 promoter in neutrophils, thus suppressing GPX4 expression and resulting in ferroptosis which contributes to disease development (Extended Data Fig. 10).

Ethics Statement

Informed consent was obtained from all human participants. Medical and economic support were provided for participants. All in vivo experiments were performed according to the guidance of the German Animal Welfare Law. Our study was approved by the Institutional Review Board and Animal Care and Use Committee of Peking Union Medical College Hospital (PUMCH) (JS-1196), Beth Israel Deaconess Medical Center’s Institutional Review Board (2006P000298) and Animal Care and Use Committee (088–2015).

Chemicals, cytokines and antibodies

Chemicals and cytokines used were as follows: recombinant human IFN α2b (11100–1, PBL assay science), RSL3 (S8155, Selleck), liproxstatin-1 (LPX-1) (S7699, Selleck), deferoxamine mesylate (DFO) (S5742, Selleck), necrostatin-1 (Nec-1) (S8037, Selleck), necrosulfonamide (NSA) (S8251, Selleck), Z-VAD-FMK (S7023, Selleck), PMA (P8139, Sigma), GSK2795039 (HY-18950, MedChem Express), APX-115 (HY-120801, MedChem Express), Cl-amidine (506282, Merck), β-ME (M6250, Sigma), dimethyl sulfoxide (DMSO) (D8371, Solarbio), cytoxan (CTX) (S2057, Selleck), BODIPY™ 581/591 C11 (D3861, Invitrogen), penicillin (Gibco), streptomycin (Gibco), lysing buffer (555899, BD Pharmingen), DAPI (S2110, Solarbio), 7AAD (559925, BD Pharmingen), propidium iodide (P1304MP, Thermo Scientific), Sytox Green (S7020, Invitrogen), Hoechst (62249, Thermo Scientific) and 16% formaldehyde (12606S, CST).

Antibodies used were as follows: anti-glutathione peroxidase 4 (GPX4) antibody (EPNCIR144, ab125066, Abcam; 1:1000), anti-CaMKIV antibody (ab3557, Abcam; 1:1000), anti-histone 3 antibody (ab1791, Abcam; 1:1000), anti-C1q antibody (4.8, ab182451, Abcam, 1:200), anti-neutrophil elastase antibody (ab68672, Abcam, 1:200), anti-CREM antibody (WB: nbp2–16009, NOVUS; 1:1000. CHIP: A5624, Abclonal), anti-GAPDH antibody (10494–1-AP, Proteintech; 1:1000), anti-β-actin (BE0022, Easybio; 1:1000), anti-SLC7A11 antibody (26864–1-AP, Proteintech; 1:1000), anti-histone H3 (citrulline R2) antibody (EPR17703, ab176843, Abcam; 1:1000), anti- FcγR3β antibody (MM0272–5L11, ab89207, Abcam; 1:1000), anti-rabbit IgG-HRP (BE0101, Easybio; 1:5000), anti-human IgG-HRP (BE0122, Easybio; 1:5000), anti-mouse IgG-HRP (BE0102, Easybio; 1:5000), PE conjugated anti-mouse Ly-6G (127607, Biolegend; 1:100), PE conjugated anti-human CD16 (561313, BD Pharmingen;1:100), APC conjugated anti-mouse/human CD11b (101212, Biolegend; 1:100), FITC conjugated anti-mouse CD45 (103108, Biolegend; 1:100), FITC conjugated anti-mouse Ly-6G&Ly-6C (553127, BD Pharmingen; 1:100), TruStain FcX anti-CD16/32 (422302, Biolegend; 1:100), goat pAb to mouse IgM Alexa Fluor 647 (ab150123, Abcam; 1:200), goat pAb to rabbit IgG Alexa Fluor 488 (ab150077, Abcam; 1:200), donkey pAb to rabbit IgG Alexa Fluor 647 (ab150075, Abcam; 1:100), goat anti-mouse IgG Alexa Fluor 594 (405326, Biolegend; 1:200), mouse monoclonal antibody against human interferon Alpha/Beta Receptor 1 (MMHAR-3, 21370–3, pbl assay science), anti-CXCL11 antibody (ab9955, Abcam), and anti-human IL-12/23 (ustekinumab) (HY-P9909, Medchem Express).

Patients and healthy controls (HCs)

SLE patients involved in this study were recruited from two cohorts, all fulfilling the diagnosis of SLE according to the criteria established by the American College of Rheumatology⁴³. Peripheral blood samples were obtained from new onset untreated patients recruited from PUMCH. SLE patients were treated with standard of care medications including corticosteroid and immunosuppressants (such as mycophenolate mofetil (MMF) or CTX) combined with hydroxychloroquine for at least 2 months, and post-treatment blood samples from these same SLE patients were collected. WBC count, neutrophil count, and anti-dsDNA antibodies were measured and analyzed on the day of blood collection. Disease activity was assessed by the modified SLEDAI-2000 calculated by omitting the immunologic variables including anti-dsDNA antibodies and serum complement concentrations from the SLEDAI-2000 ⁴⁴.

SLE patients from another cohort (Caucasian n=13, African Americans n=15) were recruited from the Division of Rheumatology, Beth Israel Deaconess Medical Center.

Rheumatoid arthritis, ankylosing spondylitis, and Behcet’s disease patients involved were recruited from PUMCH, and were all untreated and diagnosed according to the corresponding criteria^{45, 46, 47}. Patients with infections or severe underlying disorders were also excluded. HCs were age and gender-matched individuals without autoimmune, inflammatory or infectious diseases. Clinical characteristics and the treatment of SLE patients included in the study were shown in Supplementary Table 1-5 and 7.

Mice

Female MRL/Mpj and MRL.Mpj-Fas^lpr (MRL/lpr) mice were purchased from Shanghai Slack Laboratory, and maintained in the specific-pathogen-free conditions of PUMCH animal care facility.

NZB and NZW mice were purchased from Jackson lab, bred at Beijing Vital River Laboratory, and the first female generation (NZB/W F1) after crossbreeding was used for experiments.

The C57/B6-Gpx4^fl/fl (027964) and C57/B6-LysMcre (004781) mice were purchased from The Jackson Laboratory. The Gpx4^fl/fl allele has loxP sites flanking exons 2–4 of the Gpx4 gene. The LysMcre knock-in allele has a nuclear-localized Cre recombinase inserted into the first coding ATG of the lysozyme 2 gene (Lyz2) both abolishing endogenous Lyz2 gene function and placing NLS-Cre expression under the control of the endogenous Lyz2 promoter/enhancer elements. The Gpx4^fl/fl mice were bred to the LysMcre mice to generate LysMcre+ Gpx4^fl/wt versions of these strains. Gpx4^fl/fl and LysMcre⁺ Gpx4^fl/wt mice were used for experiments. All mice were bred and maintained in the animal facility of Beijing Vital River Laboratory under specific-pathogen-free conditions.

Female Camk4^−/− C57BL/6 mice were obtained from The Jackson Laboratory. Female Crem^−/− C57BL/6 and Camk4^−/− MRL/lpr, mice are generated and maintained in the specific-pathogen–free animal facility at Beth Israel Deaconess Medical Center, Harvard Medical School. In all experiments, mice were carefully matched for age, sex, weight, and genetic background.

In-vivo experiments

In ferroptosis rescue experiments, LPX-1 (10 mg/kg), CTX (20 mg/kg) or 0.1ml DMSO (10%) was administered intraperitoneally to female MRL/Mpj and MRL/lpr mice every other day for six weeks beginning at 12 weeks of age. Urine was collected in the morning once a week. Mice were sacrificed at 18 weeks of age, and spleen, kidneys, lymph nodes, and peripheral blood were collected for analysis. For in vivo treatment, MRL/lpr mice were administered 0.1 ml DMSO (10%), Cl-amidine (20 mg/kg), LPX-1 (10 mg/kg), or Cl-amidine (20 mg/kg) combined with LPX-1 (10 mg/kg) intraperitoneally every other day for 3 weeks starting at 12 weeks of age⁴⁸. Mice were euthanized at 15 weeks of age and spleen, lymph nodes, urine and peripheral blood were collected for analysis.

Peripheral blood and bone marrow were collected from three- or seven-month-old NZB/W F1, Gpx4^fl/fl and Gpx4^fl/wt LysMcre⁺, MRL/Mpj and MRL/lpr mice. Neutrophils, lymphocytes and monocytes were isolated with mouse peripheral blood lymphocyte separation kit (P8620, Solarbio), mouse bone marrow neutrophil isolation kit (P8550, Solarbio), mouse peripheral blood neutrophil separation kit (P9201, Solarbio) and mouse peripheral blood monocyte separation kit (P5230, Solarbio)

Gpx4^fl/fl and Gpx4^fl/wt LysMcre⁺ mice were observed without interventions. Urine was collected in the morning once a month, and tail venous blood was collected for analysis. Absolute neutrophil counts were analyzed by flow cytometry. Neutrophils were isolated and cultured in complete medium supplemented with LPX-1 (1 μM) for 16 hours and assessed for cell viability and lipid-ROS.

Pristane-induced murine lupus: 2 months old female Wild type, Crem^−/− and Camk4^−/− C57BL/6 were i.p. administered with Pristane (0.5 ml/mouse, P2870, Sigma). After 2 months, 2 × 10^9 pfu Adenovirus IFNα per mouse were i.v. administered and mice were euthanized 1 month later.

Cell isolation and in vitro culture

To isolate neutrophils and peripheral blood mononuclear cells (PBMCs), blood was layered on Ficoll-Hypaque density gradients and centrifuged following the manufacturer’s instructions. Erythrocytes were lysed with Lysing Buffer. Monocytes, B cells and plasmacytoid dendritic cells (pDC) were purified with CD14 microbeads (130050201, Miltenyi), CD19 microbeads (130050301, Miltenyi) and plasmacytoid dendritic cells isolation kit II human (130097415, Miltenyi) respectively. The preparation contained greater than 98% neutrophils as confirmed by flow cytometry using anti-CD16 and anti-CD11b antibodies. Cells were cultured in complete RPMI 1640 basic medium (Gibco) with 100 U/mL penicillin, 100 μg/mL streptomycin, and 20% serum from 10 randomly selected SLE patients or age-and gender-matched HCs, incubated at 37℃ in a humidified atmosphere of 20% O₂ and 5% CO₂. IFN-α2b (10³-10⁵ U/ml), anti-IFNAR1 (0.1–10 μg/ml), anti-CXCL11 antibody (0.1–10μg/ml), Ustekinumab (0.1–10 μg/ml), RSL-3 (10 μM), LPX-1 ( 0.01–1 μM), DFO ( 1–100 μM), Nec-1 ( 0.01–1 μM), NSA ( 0.01–1 μM), Z-VAD (0.01–1 μM), β-ME (10 or 50 μM), PMA (50nM), Cl-amidine (100 μM), GSK2795039 (10 μM), or APX-115 (5 μM) were added in corresponding experiments. Neutrophils were cultured for 4–30 hours to further analysis. Cultural supernatant of pDC and B cells was collected at 24h and 72h respectively.

In relevant experiments, purified SLE IgG (1.2–3.6 g/L) was added to complete medium.

Quantification of NET formation by Fluorescence Microscope

HC neutrophils were cultured in complete RPMI medium supplemented with 20% HC or SLE serum in the presence or absence of liproxstatin-1 (1 μM), Cl-amidine (100 μM), IFN-α2b (10⁵ U/ml), SLE IgG (3.6g/L) for 4 or 16 hours at a density of 1 × 10⁵ cells/ml in 96-well plates. The release of NET-DNA was measured with DAPI and propidium iodide or Sytox Green and Hoechst by Fluorescence Microscope (ZEISS). Neutrophils whose DNA area exceeds 400 μm² were considered undergoing NETosis as previously described⁴⁹.

Cell viability quantification

The viability of freshly isolated or cultured neutrophils, monocytes and lymphocytes was determined by the Annexin V/7-AAD assay with the PE Annexin V Apoptosis Detection Kit I (559763, BD Pharmingen). Annexin V (-) and 7AAD (-) cells were considered living cells. Lactate dehydrogenase (LDH) release was measured with Cytotoxicity Detection Kit according to manufacturer’s instructions (11644793001, Merck). The viability quantification by fluorescence microscope was similar to NET quantification, measured with Sytox Green and Hoechst.

IgG purification

SLE or HC IgG was purified using the ProteoExtract Albumin/IgG Removal Kit (122643, Merck) or Protein A/G (BE6868, Easybio) according to the manufacturer’s protocols. For kit-isolated IgG, 0.75M NaHCO3 was used to neutralize the pH of the eluent. IgG was enriched by centrifugal filters (UFC905096, Merck) and resuspended in PBS.

Transmission Electron Microscopy

Neutrophils freshly isolated or cultured with vehicle (DMSO) or RSL-3 (10 μM) for 4 hours were fixed with 2.5% glutaraldehyde in 0.1 M Sorenson’s buffer for 3 hours, and then treated with 1% OsO4 in 0.1 M Sorenson’s buffer for 2 hours. Enblock staining used 1% tannic acid. After dehydration through an ethanol series, neutrophils were embedded in Lx-112 (Ladd Research Industries) and Embed-812 (EMS). The thin cut sections were stained with 1% uranyl acetate and 0.4% lead citrate and examined under the TEM-1400 plus electron microscope. Electron micrographs were taken at 5,000–50,000-fold magnification. 8 SLE patients and 6 HCs were evaluated for mitochondrial vacuolization by EM. We randomly collected 10 neutrophils per sample under the lens of 50,000 times and counted the number of mitochondria with vacuoles for quantitative analysis. Images were optimized using Photoshop CC 2017 and Illustrator CC 2017 (Adobe).

Flow cytometry

For neutrophil percentages measurement, whole blood cells of mice were stained with combinations of antibodies including anti-Ly-6G, anti-CD11b and anti-CD45 for 30 minutes on ice in staining buffer. For detection of intracellular GPX4 expression, cells were stained with anti- Ly-6G&Ly-6C and anti- GPX4 antibody after fixation and permeabilization with Cytofix/Cytoperm solution (554714, BD Pharmingen). After incubation for 2 hours, cells were washed before being stained with anti-rabbit IgG Alexa Fluor 647. For the lipid ROS detection, neutrophils were washed twice and incubated for 15 min with 2 μM C11-BODIPY (581/591). Afterward, neutrophils were washed twice and resuspended in PBS. Lipid-ROS was assessed by measuring the fluorescence change of C11-BODIPY at 488nm using BD FACS Aria II (BD, Biosciences, CA, USA). Data was analyzed using FlowJo (version 10.4, Tree Star).

Multiplex cytokine detection

Cytokines and immunoglobulin in human and mouse plasma were measured with the LEGENDplexTM mouse Th cytokine panel (740005, Biolegend), mouse immunoglobulin isotyping panel (740493, Biolegend), human cytokine panel 2 (740102, Biolegend), human Th cytokine panel (740721, Biolegend), and human proinflammatory chemokine panel (740003, Biolegend) following the manufacturers’ instructions. Data analysis was carried out using LEGENDplex™ Data Analysis Software (version 8.0). Heatmaps were constructed using the Graphpad Prism 7, and Venn diagrams completed on http://bioinformatics.psb.ugent.be/.

Enzyme-linked immunosorbent assay (ELISA)

Mouse serum was diluted 1:100 with assay buffer for anti-dsDNA antibody detection and 1:25000 for complement 3 detections. Assays were carried out with Mouse Anti-dsDNA IgG ELISA Kit (5120, Alpha Diagnostic Intl. Inc.) and Complement 3 ELISA Kit (6270, Alpha Diagnostic Intl. Inc.), following the manufacturer’s protocols. Urine protein concentrations were detected using BCA Protein Assay Kit (23225, Thermo Scientific). IgG of B cell cultural supernatant was measured by Human IgG ELISA Kit (E88–104, Bethyl).

Western blot and nuclear separation

Cells were lysed on ice using RIPA buffer (Huaxingbio) supplemented with a protease inhibitor and phosphatase inhibitor cocktail (78446, Thermo Scientific). Protein concentration of lysates was determined using the BCA Protein Assay Kitaccording to manufacturers’ instructions. Cell lysates were boiled for 10 min at 95°C with SDS, and subjected to 12% SDS–PAGE, and then transferred to PVDF membranes (1620177, Bio-Rad). The membranes were blocked and then incubated with anti-GPX4, anti-CREM, anti-CaMKIV, anti-SLC7A11, anti- FcγR3β, anti-Histone 3, anti-GAPDH or anti-β-actin antibodies overnight at 4℃. The membranes were washed and incubated with anti-rabbit or -mouse IgG-HRP for 1h. Protein bands were visualized with the western blotting detection system Tanon-5200 (Bio-Tanon, China). Gray value analysis was done by ImageJ (version 1.50g, NIH) software. Nuclear content of cells was isolated and determined by Cytoplasmic & Nuclear Extraction Kits (SC-003, Invent) according to manufacturer’s instructions.

Immunofluorescence

For NET staining, freshly isolated neutrophils (10⁵ cells) from HC and SLE patients were transferred into a 24- well plate with BioCoat poly-L-lysine slides (Corning), and cultured in complete RPMI 1640 basic medium with 20% HC or SLE serum in the presence or absence of liproxstatin-1 (1 μM) and PMA (50 nM). After incubation at 37℃ for 4 hours, cells were fixed with 4% paraformaldehyde. NETs were stained using anti-neutrophil elastase antibody for 2 hours, and then anti-rabbit IgG Alexa Fluor 488. Slides were mounted with anti-quenching agent containing DAPI.

Kidneys were taken from mice, fixed in 4% paraformaldehyde, embedded in paraffin and stained with Periodic Acid-Schiff (PAS). The second kidney from each mouse was frozen in O.C.T compound (Tissue-Tek) at −80℃, and used for assessment of immune complex deposition. 6-μm sections of snap-frozen kidney tissue were treated with blocking buffer (100 mM Tris-HCl, pH 8.0, 0.3% Triton X-100, 2% BSA, and 50 μg/ml goat non-specific IgG) for 1 hour. Afterward, the tissue was stained with anti-mouse IgG Alexa Fluor 594, anti-mouse IgM Alexa Fluor 647, or anti-C1q antibody plus anti-rabbit IgG Alexa Fluor 488. Coverslips were mounted in anti-quenching agent with DAPI. All slides were examined with the A1 HD25/A1R HD25 Nikon confocal laser microscopy (Nikon, Japan). Images were optimized using Photoshop CC 2017 and Illustrator CC 2017 (Adobe).

RNA isolation and qPCR

Total RNA was extracted with TRIzol (10296010, Invitrogen), and converted to complementary DNA using PrimeScript™ RT Master Mix (RR036A, Takara Bio). Real-time quantitative PCR was carried out with the Applied Biosystems 7500 using SYBR Premix Ex TaqTM Ⅱ (RR820A, Takara Bio). The relative RNA expression level was normalized to GAPDH mRNA according to the △△Ct calculation method.

The primers used were as follows:

GAPDH

Forward:5’- TCAACGACCACTTTGTCAAGCTCA-3’

Reverse: 5’- GCTGGTGGTCCAGGGGTCTTACT-3’,

GPX4

Forward:5’- AAGTGGATGAAGATCCAACCCAAG- 3’

Reverse: 5’- GGGGCAGGTCCTTCTCTATCA- 3’,

CREM

Forward:5’- CGTCGACATTCTTTGGCAGC-3’,

Reverse: 5’- ATGACCATGGAAACAGTTGAATCCCA-3’,

ISG15

Forward:5’- TGGACAAATGCGACGAACCTC -3’,

Reverse: 5’- TCAGCCGTACCTCGTAGGTG -3’,

IFIT1

Forward:5’- GCGCTGGGTATGCGATCTC-3’,

Reverse: 5’- CAGCCTGCCTTAGGGGAAG-3’

RNA-seq analysis

Total RNA of freshly isolated HC or SLE neutrophils was extracted with TRIzol, and subjected to RNA-seq analysis. RNA sequencing was performed by the Novogene Experimental Department using Illumina Hiseq 4000 platform. Raw reads of fastq format were firstly processed through in-house perl scripts and all the downstream analyses were based on the clean data with high quality. Index of the reference genome was built using bowtie2 v2.2.8 and paired-end clean reads were aligned to the reference genome using HISAT2 v2.0.4. The mapped reads of each sample were assembled by StringTie (v1.3.1). Cuffdiff (v2.1.1) was used to calculate FPKMs of coding genes in each sample and provided statistical routines for determining differential expression using a model based on the negative binomial distribution. Transcripts with a P-adjust <0.05 were assigned as differentially expressed.

DNA pull-down

Nuclear lysates were obtained from HC or SLE neutrophils as described above. Biotin-labeled Gpx4 promotor DNA was synthesized by Guangzhou RiboBio (Forward: 5’- ATTGGCTGACGTCG-3’, Reverse: 5’- CGACGTCAGCCAAT -3’). The biotinylated promotor DNA oligos were annealed and incubated with nuclear lysates at room temperature for 1 hour in binding buffer (10 mM Tris, 1 mM KCl, 1% NP-40, 1 mM EDTA, 5% glycerol). Afterward, M-280 streptavidin Dynabeads (10004D, Invitrogen) were added and incubated for 2 hours at 4°C with rotation. After three washes with binding buffer, the GPX4-promotor-binding proteins were eluted by boiling and analyzed by western blot.

Chromatin Immunoprecipitation (CHIP) Analysis

CHIP analysis was conducted on neutrophils with anti-CREM (Abclonal) and the assay was performed according to the manufacturer’s instructions (9003, CST). Each sample (4 × 10⁶ cells) was cross-linked with 1% formaldehyde and then subjected to nuclear extraction and chromatin digestion with micrococcal nuclease. Sonication was used for complete lysis of nuclei (10-s pulse and 30-s pause; five cycles). The efficiency of chromatin digestion was determined by nucleic acid electrophoresis. For immunoprecipitation, digested chromatin was incubated with 10 μg of antibodies for 6 hours at 4 °C with rotation. Afterward, magnetic beads were added to the immunoprecipitation reaction for 2 hours at 4 °C with rotation. After washing four times, immunoprecipitated chromatin DNA was eluted. Fold enrichment was quantified using qPCR with SYBR Green Realtime PCR Master Mix (QPK-201, TOYOBO) and calculated as a percentage of input chromatin (% input). The primers sequence of Gpx4 promotor was 5’- AACAAGTCCGCACGTCCGGT-3’(forward) and 5’- ATTGGTCAGACGCGTCGGTGTT-3’(reverse).

RNA interference and plasmids

HL60 cells (KG141, Keygen) were transfected with siRNA (20nmol/ml) or plasmids (1μg/ml) of Crem (NCBI Gene ID:1390) and Fcγr3b (NCBI Gene ID: 2215). 293T cells (CL-0005, Procell) were transfected with plasmids (1μg/ml) of Slc7a11(NCBI Gene ID: 23657). Full-length sequences were obtained from Sangon Biotech and then cloned into pcDNA3.1 vector. For transfection, cells were treated with the Gene Pulser II electroporation system (Bio-Rad) or 4D-Nucleofector X Kit L (V4XP-3024, LONZA,) or lipo-2000 (11668–027, Invitrogen) according to the manufacturer’s instructions. The efficiency of interference or overexpression was determined by qPCR or Western blot. The corresponding siRNAs used are as follow:

CREM:

siRNA-1

Forward:5’- GCAGAAUCAGAAGGUGUAATT-3’

Reverse: 5’- UUACACCUUCUGAUUCUGCTT-3’,

siRNA-2

Forward:5’- GGUGGAACAAUCCAGAUUUTT-3’

Reverse: 5’- AAAUCUGGAUUGUUCCACCTT-3’,

siRNA-3

Forward:5’- CCCAGGAUCUGAUGGUGUUTT-3’

Reverse: 5’- AACACCAUCAGAUCCUGGGTT-3’,

Statistics and Reproducibility

Each experiment was repeated independently more than three times, and similar results were obtained. All data were analyzed using GraphPad Prism7 software. Shapiro-Wilk test was used for normal distribution verification. For data with a normal distribution and homogeneity of variance, the independent sample t-test was used to compare differences between two groups. In some experiments, a paired Student’s t-test was applied as indicated. For data with non-normal distribution, Mann Whitney test were applied. Correlations were calculated using Pearson rank correlation analysis. Sample sizes were determined on the basis of previous experiments using similar methodologies and were detailed in each figure legend. For in vivo studies, mice were randomly assigned to treatment groups. Data in figure legends are presented as mean ± SD values or median with interquartile range. P<0.05 was considered significant.