Growth and differentiation factor 15 (GDF15), a divergent member of the transforming growth factor-β superfamily, has been associated with acute and chronic inflammatory conditions including autoimmune disease, i.e., type I diabetes and rheumatoid arthritis. Still, its role in systemic autoimmune disease remains elusive. Thus, we studied GDF15-deficient animals in Fas-receptor intact (C57BL/6) or deficient (C57BL/6lpr/lpr) backgrounds. Further, lupus nephritis (LN) microdissected kidney biopsy specimens were analyzed to assess the involvement of GDF15 in human disease. GDF15-deficiency in lupus-prone mice promoted lymphoproliferation, T-, B- and plasma cell-expansion, a type I interferon signature, and increased serum levels of anti-DNA autoantibodies. Accelerated systemic inflammation was found in association with a relatively mild renal phenotype. Splenocytes of phenotypically overall-normal Gdf15−/− C57BL/6 and lupus-prone C57BL/6lpr/lpr mice displayed increased in vitro lymphoproliferative responses or interferon-dependent transcription factor induction in response to the toll-like-receptor (TLR)-9 ligand CpG, or the TLR-7 ligand Imiquimod, respectively. In human LN, GDF15 expression was downregulated whereas type I interferon expression was upregulated in glomerular- and tubular-compartments versus living donor controls. These findings demonstrate that GDF15 regulates lupus-like autoimmunity by suppressing lymphocyte-proliferation and -activation. Further, the data indicate a negative regulatory role for GDF15 on TLR-7 and -9 driven type I interferon signaling in effector cells of the innate immune system.
Systemic lupus erythematosus (SLE) is a classic example of systemic autoimmunity. Genetic susceptibility, female sex-specific factors, and environmental triggers, e.g., viral infections are key determinants of SLE onset. In established disease, a misguided “self-directed” adaptive immune response provokes autoantibody production and lymphadenopathy. Furthermore, erroneous recognition of nucleic acids, e.g., by toll-like receptors (TLRs) and the subsequent type I interferons secretion by dendritic cells and macrophages confirm an essential role of innate immunity in the pathogenesis of SLE . Both systems are involved in organ damage resulting from autoimmunity. For instance, in lupus nephritis (LN) immune complex deposition induces local immune cell activation and increases immune cell influx leading to tissue damage [1, 2].
Growth differentiation factor 15 (GDF15), originally discovered as macrophage inhibitory cytokine 1 (MIC1), is a cytokine belonging to the transforming growth factor (TGF-β) superfamily [3, 4]. Although the members of this family are structurally related, they are quite divergent from each other regarding their amino acid sequences and their specific activities in vitro and in vivo . GDF15 shares only 30% sequence homology with other family members such as TGF-β1 suggesting a nonredundant biologic activity [3, 6]. Nevertheless, the latter is well known for its immune-regulatory functions. Moreover, TGF-β1 deficiency leads to a lethal autoinflammatory disorder with some autoimmune features in mice [7, 8].
The constitutive expression of GDF15 is restricted to reproductive organs whereas other cell types, e.g., macrophages, show inducible expression upon tissue injury or pro-inflammatory, stress-induced factors such as TNFα . For instance, in LPS-induced murine sepsis, GDF15 improved survival by stimulating hepatic triglyceride metabolism to promote tissue tolerance against inflammatory damage . Studies by Kempf et al.  have demonstrated the anti-apoptotic function of GDF15 in response to heart ischemia and myocardial infarction. More recently, their studies proved that the lack of GDF15 upturns the infiltration of polymorphonuclear leukocytes into the site of inflammation. Consequently, GDF15 reduces leukocyte recruitment by abating chemokine signaling . Our recent study suggests the modulation of the CXCR3 receptor in activated T-cells as a possible GDF15-dependent mechanism that regulates the infiltration of T cells in anti-GBM nephritis in mice .
How GDF15 modulates immune-cell signaling is not clear yet. Expression of its known receptor GDNF family receptor alpha-like (GFRAL) is restricted to cerebral tissues and additional signaling pathways seem mandatory . Most recently, an immunosuppressive effect via CD48/STUB1 signaling in Tregs has been reported in a hepatocellular carcinoma transgenic mouse model . In accordance, others found the GDF15 to stimulate Treg maturation with a significant correlation between FoxP3 and GDF15 expression . GDF15 was reported to suppress dendritic cells (DCs) maturation, DC-MHCII-expression, and IL-12 secretion in tumor-bearing mice . GDF15 was also shown to act as a paracrine factor, required for M2-like macrophage polarization, reversal of inflammatory adipose tissue inflammation and insulin resistance in mice . Cumulatively, these data indicate that the immune regulatory role of GDF15 goes beyond acute inflammation and tissue injury. GDF15 might impact chronic inflammation and autoimmunity. In fact, GDF15 serum levels were shown to be positively correlated with disease activity in rheumatoid arthritis patients [9, 18]. In addition, proteomic analysis has shown an absence of GDF15 in islet cells from individuals with an inflammatory stage of type I diabetes . Likewise, LN patients from a decent-sized cohort showed elevated serum GDF15 levels. Although only a mild association (r = 0.23, p = 0.03) of GDF15 with SLEDAI-2K clinical activity scoring was reported, the authors found a strong univariate association of serum GDF15 with creatinine levels (r = 0.71, p < 0.001) and proteinuria (r = 0.56, p < 0.01) .
Considering (a) the ability of GDF15 to stimulate immunoregulatory Treg maturation as well as their role in immune tolerance  and (b) the pleiotropic suppressive function of GDF15 on DC and macrophage activation and their role in systemic inflammation , we hypothesized that GDF15 suppresses progression of lupus-like autoimmunity in mice. In particular, we postulated an aggravated serologic and nephritic phenotype in lupus-prone C57BL/6lpr/lpr mice in the absence of GDF15.
Materials and Methods
Mouse colonies were generated by backcrossing the Gdf15−/− strain (MGI: 2386300, Gdf15tm1Sjl)  into C57BL/6J parental strain obtained from Jackson Laboratories for at least 6 generations. Later Gdf15−/−lpr/lpr mice were generated by crossing Gdf15−/− and C57BL/6lpr/lpr (B6.MRL-Faslpr/J, Stock No: 000482 obtained from Jackson Laboratories) for four generations. Newly established colonies of wild-type and mutant mice in the C57BL/6 background (parental strain C57BL/6J) were used for experiments. The Gdf15−/−lpr/lpr mice were born at Mendelian ratios and display common litter sizes. The genotype was assured by PCR. Mice were housed in groups of 5 mice in sterile filter top cages with a 12 h dark/light cycle and unlimited access to autoclaved food and water. We used only female mice for all experiments. All mice were sacrificed by cervical dislocation at 24 weeks of age. This study was carried out following the principles of the Directive 2010/63/EU on the Protection of Animals Used for Scientific Purpose and with approval by the local government authorities.
Evaluation of Autoimmune Tissue Injury
Organs (spleens and kidneys) from female mice were fixed in 4% buffered formalin, processed, and embedded in paraffin. The severity of the renal lesions was graded on periodic acid-Schiff (PAS) stained sections using the indices for activity and chronicity. Histology stainings were performed following the manufacturer’s instructions. We used following antibodies: rat Blimp1 (cat. 14-5963-82, Invitrogen, 1:100); rat B220 (cat. 14-0452-82, Invitrogen, 1:400); Ki67 (Bio-Rad, 1:100); Mac2 (125402 BioLegend, 1:3000); CD45 (cat. 5188647, BD, 1:100); C9 (PA5-29093, Invitrogen, 1:50); WT-1 (MA5-32215, Invitrogen, 1:100). For quantitative analysis, glomerular cells were counted in 10 cortical glomeruli per section (from “n” mice in the group, as indicated in a figure) or were analyzed using Adobe Photoshop CS4Extended (Adobe, San Jose, CA, USA) (% of stained high power field). The kidney function of 6-month-old female mice from every group was determined by measuring serum creatinine levels determined by the Jaffe method (DiaSys Diagnostic Systems). Albuminuria and proteinuria were determined via mouse Albumin Quantification Set (Bethyl Laboratories, Montgomery, TX, USA) and Bradford method, respectively. Urinary albumin excretion was evaluated by a double-sandwich ELISA. First, 96-well plates were coated with goat anti-mouse albumin antibody A90-13A-5 (Bethyl Laboratories, Montgomery, TX, USA), and plates were incubated overnight at 4°C. After blocking for half an hour at room temperature in 0.5% BSA in PBS with 0.05% Tween20, urine samples, and mouse albumin standard (Sigma-Aldrich, St. Louis, MI, USA) were added on the plate in triplicates for 2 h. Mouse urine samples were diluted in serial dilutions ranging from 1:102 to 1:107. As a secondary antibody, HRP-conjugated anti-mouse albumin antibody A90-134P-7 (Bethyl Laboratories) was used.
Compensation was done using the following antibodies on pooled mice splenocytes. Before gating doublets were excluded on FSC-(H)ight versus FSC-(W)idth blots. After compensation using the following antibodies on pooled mice splenocytes, splenic T cells were identified using anti-mouse CD3 FITC (#553062 BD), CD4 APC (#553051 BD), CD8a PerCP (#553030 BD), and CD44 P3 (#553134 BD). PI was used for dead cell exclusion. Tregs were identified using the BD Th17/Treg mouse phenotyping kit following the manufacturer’s instructions (#560767 BD). B cells were stained using anti-mouse FITC MHC II (11532185 eBioscience), B220 Alexa Fluor 647 (clone RA3-6B2 BioLegend) and CD19 PB450 (clone 6D5, BioLegend). PI was used for the exclusion of dead cells. Plasma cells were stained intracellularly with anti-mouse kappa light chain PE (#559940, BD). In addition, anti-mouse CD138 APC (#558626, BD) and anti-mouse FITC MHC II (#11532185 eBioscience) were used. Zombie NIRTM fixable viability dye (#423106 BioLegend) was used for dead cell exclusion. Splenic macrophages and dendritic cells were analysed after staining for anti-mouse CD86 FITC (#553691 BD), CD11c PE (#557401, BD), F4/80 APC (clone MCA497 BioRad/AbDSeroTec). PI was used for dead cell exclusion. Plasmacytoid dendritic cells (pDCs) were identified using CD11c PE (#557401, BD), B220 Alexa Fluor 647 (clone RA3-6B2 BioLegend), and anti-mouse CD317 PE (# 127104, BioLegend). Intracellular labeling was done using the Cytofix/Cytoperm kit (BD), following the manufacturer’s instructions. Cell counting beads (Invitrogen) were used for determining cell numbers by FACS Cytoflex (Beckman Coulter, Indianapolis, IN, USA) and analyzed with FlowJo V.10 Software (BD). The gating strategy is presented in online supplementary Figure 4 (see www.karger.com/doi/10.1159/000523991 for all online suppl. material).
Real-Time Quantitative PCR
Real-time RT-PCR was performed on total spleen mRNA and whole blood samples as well as in vitro experiments. Shortly SYBR Green Dye detection system was used for quantitative real-time PCR on Light Cycler 480 (Roche, Mannheim, Germany). Gene-specific primers (225 nM, Metabion, Martinsried, Germany) were used as listed in Table 1. Controls consisting of ddH2O were negative for target and housekeeper genes. Gapdh mRNA was used as a housekeeper. The PCRs were performed in 96-well plates according to the manufacturer protocol (Light Cycler 480 II). Standard controls on each array for genomic DNA contamination, RNA quality, and general PCR performance were included.
Autoantibody and Plasma Analysis
For detection of antibodies to double-stranded DNA, NUNC MaxiSorp ELISA plates (Thermo Fisher Scientific) were coated with poly-L-lysine (Trevigen) and PBS (ratio 1:1) for 1 h. Plates were washed with TrisNaCl (50 mM Tris and 0.14 M NaCl pH 7.5) and double-stranded DNA (2 μg/mL) was coated in SSC pH 7.0 buffer overnight. For anti-Smith and anti-histone antibodies, MaxiSorp ELISA plates were coated with Smith antigen (Sm3000; Immunovision) or histones (2 μg/mL) in 0.05 M carbonate-bicarbonate buffer overnight at 4°C. Serum samples were diluted 1:100–1:1.000.000 for all IgM/IgG ELISAs. Sera from 24-week C57BL/6, C57BL/6lpr/lpr, 24-week MRLlpr/lpr mice were used as controls. Horseradish peroxidase-conjugated antibody to mouse IgG and IgM (A90-131P and A90-101A, Bethyl) was used as a secondary antibody (1:50.000). Absorbance was measured at 450 nm with a Sun-rise plate reader (TECAN). BD Cytometric Bead Array (CBA) Mouse Inflammation Kit as well as standard BD and R&D ELISAs (Mouse Quantikine ELISA Kits) were used to determine the pro-inflammatory cytokines according to manufacturer protocol. For cytokine analysis of serum from mice or cell culture experiments, samples were prepared according to the instruction of the BD Cytometric Bead Array Mouse Inflammation Kit. The concentrations of the cytokines IL-6, IL-10, MCP-1, IFN-γ, TNF, and IL-12p70 in the samples were determined by the software FCAP software V2. Plasma aldosterone levels were analysed using an enzyme-linked immunosorbent assay (ELISA, RE52301, IBL International GmbH, Hamburg, Germany). Briefly, the plasma samples were diluted (1:5) with deionized water and loaded into the microtiter wells coated with polyclonal rabbit antibodies directed against the antigenic site of the aldosterone molecule. All consequent steps were performed as per the manufacturer`s protocol. The absorbance (OD) of each well was measured at 450 ± 10 nm using a microtiter plate reader (FLUOstar Omega, Serial no. 415-2731, BMG Labtech, Ortenberg, Germany). Plasma corticosterone levels were measured using a high sensitivity enzyme immunoassay (EIA, AC-15F1, Immunodiagnostic Systems Limited, United Kingdom) as per the manufacturer’s protocol. Briefly, the plasma samples were diluted with the provided diluent in the ratio of 1:50 and heated at 80°C for 10 min. The samples were then incubated with horseradish peroxidase (HRP)-conjugated polyclonal corticosterone antibody in the microtiter wells for 4 h at room temperature. The wells were washed and TMB (3,3′,5,5′-Tetramethylbenzidine, chromogenic substrate) was added. The generated color was measured at the absorbance of 450 nm in a microplate plate reader (FLUOstar Omega, Serial no. 415-2731, BMG Labtech, Ortenberg, Germany). The amount of extracellular DNA in serum was quantified using Quant-iT PicoGreen dsDNA Reagent (QPG; Invitrogen/ThermoFisher Scientific). QPG was diluted 1:200 in TE buffer (10 mM Tris, 1 mM EDTA [pH 7.5]), and 90 μL was mixed with 10 μL of serum (dilution 1:2–1:50) containing extracellular DNA. The fluorescence was measured at an excitation wavelength of 480 nm and an emission wavelength of 520 nm. We used the information about numbers of long interspersed nuclear element-1 (LINE-1 or L1) elements in mammalian genomes to prepare the standard curve ; 1 μg of mouse genomic DNA corresponds to 3.4 × 105 copies of a single-copy gene. We used following primers to estimate the cell-free DNA in 2 μL (1:10 diluted) serum (mitochondrial DNA Fw: GCCCATGACCAACATAACTG, Rv: CCTTGACGGCTATGTTGATG; Line element Fw: ACCAAATGGCTGAGAAGCAC, Rv: ATCTGCTGTCGCCTGAATTT).
In vitro Experiments
Splenocytes cells from wild-type and knockout mice were isolated from fresh spleens, erythrocyte-depleted, and cultured in DMEM medium supplemented with 2% FCS and 1% penicillin/streptomycin for 48 h. Bone marrow cells from wild-type and knockout mice were isolated from femur and tibia, erythrocyte-depleted, and cultured with 20 ng/mL mouse recombinant M-CSF (Immunotools, Friesoythe, Germany) in DMEM medium supplemented with 10% FCS and 1% penicillin/streptomycin for 7 days to generate macrophages. Bone marrow cells from wild-type and knockout mice were isolated from femur and tibia, erythrocyte-depleted, and cultured with 20 ng/mL mouse recombinant GM-CSF or Flt3L (Immunotools, Friesoythe, Germany) in DMEM medium supplemented with 10% FCS and 1% penicillin/streptomycin for 7 days to generate cDCs and pDCs, respectively. On day 7, cells were stimulated with 10 ng/mL LPS, 100 ng/mL ODN 1668 B-class CpG specific for mouse TLR-9, 1 μg/mL Imiquimod, 100 ng of Poly(I:C), 20 ng/mL IFNγ, or left untreated as media control. All ligands and cytokines were obtained from Invivogen. After 4 h, macrophages were lysed and processed for gene expression analysis (qRT-PCR). The cytokines in the supernatants were determined after 24 h. Cell viability and metabolic activity were assessed by MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay accordingly to manufacturer instruction (Sigma-Aldrich). We used 50 ng/mL ODN 1668 B-class CpG specific for mouse TLR-9. Mouse splenocytes from WT and knockout mice were assessed by MTT assay up to 72 h in medium supplemented with 2% FCS; 10% FCS was used as an additional control.
Human Array Data
Human kidney biopsy specimens and Affymetrix microarray expression data were procured within the framework of the European Renal cDNA Bank-Kröner-Fresenius Biopsy Bank. Biopsies were obtained from patients after informed consent and with the approval of the local Ethics Committees . Biopsies were processed as previously reported . Published gene expression profiles (Affymetrix GeneChip Human Genome U133A and U133 Plus2.0 Arrays; GSE99340, GSE32591, GSE35489, GSE37463) used in this study came from patients with HTN(Glom: n = 15; Tub: n = 21), minimal change disease (MCD; Glom: n = 14; Tub: n = 15), and LN (Glom: n = 32; Tub: n = 32). Pretransplantation kidney biopsies from living donors (LD, Glom: n = 42, Tub: n = 42) were used as control renal tissue. CEL file normalization was performed with the Robust Multichip Average method using RMAExpress (Version 1.0.5) and the human Entrez-Gene custom CDF annotation from Brain Array version 18 (http://brainarray.mbni.med.umich.edu/Brainarray/Database/CustomCDF/CDF_download.asp).
The log-transformed dataset was corrected for batch effect using ComBat from the GenePattern pipeline (http://www.broadinstitute.org/cancer/software/genepattern/). To identify differentially expressed genes, the Significance Analysis of Microarrays (SAM) method was applied using TiGR (MeV, Version 4.8.1) . Tubulointerstitial and glomerular gene expression profiles from patients with different chronic kidney diseases (MCD, HTN, LN) and controls (LD) were used to compute the correlation of the log-transformed steady-state expression levels of GF15 with Type-I interferon- and Ki-67 gene expression using Spearman correlation (GraphPad Prism 8.0).
Data were expressed as mean ± SEM to take into account both the value of the SD and the sample size instead of variability among replicates. The one-way analysis of variance (ANOVA) was used to determine statistics in the case of three or more independent (unrelated) groups. Tukey’s multiple comparison test was used for all-possible pairwise comparisons. The student’s t test was used for direct comparisons between single groups, i.e., wild type and knockout cells/mice in case of normally distributed data or sample size n > 15. Mann-Whitney U test was used to analyze data with small sample size and nonparametric distribution of data. We used GraphPad Prism software. Statistical significance was indicated as follows: p value of <0.05 (*); p value of <0.01 (**); p value of <0.001 (***).
GDF15 Suppresses Lymphocyte Proliferation in Murine Lupus-Like Disease
First, we evaluated female GDF15-deficient mice for signs of spontaneous autoimmunity. We did not observe any significant differences between Gdf15−/− and wild-type C57BL/6 mice in body weight (Fig. 1a). Up to 6 months of age Gdf15−/− mice did not display lymphoproliferation, splenomegaly, or autoantibodies in serum (Fig. 1b, c; the level of antibodies will be discussed later in the results chapter and Fig. 3). Simultaneously, we generated GDF15-deficient C57BL/6lpr/lpr mice by crossing Gdf15−/− mice with autoimmune-prone C57BL/6lpr/lpr mice deficient for the Fas-receptor (CD95) as described in materials and methods. GDF15 serum levels of 6-months old mice were comparable in both C57BL/6 and autoimmune C57BL/6lpr/lpr background (Fig. 1d). Despite similar body weights (Fig. 1a), Gdf15−/−lpr/lpr mice displayed significant lymphadenopathy and splenomegaly compared with C57BL/6lpr/lpr controls (Fig. 1b, c). Splenic Ki-67 staining revealed increased lymphocyte proliferation (Fig. 1e). Gdf15−/−lpr/lpr mice also showed increased splenic mRNA expression of DNAse-I and Caspase-3 and increased absolute numbers of late apoptotic/necrotic (PI+) splenocytes, cumulatively indicating an increased “cellular turnover” and possibly nucleic acid exposure (Fig. 1f–h). In fact, sera from Gdf15−/−lpr/lpr but not Gdf15−/− mice contained increased concentrations of cell-free DNA (Fig. 1i, j).
GDF15 Suppresses Proliferation and Activation of B- and T-Cells in lpr Mice
Splenic mRNA analysis suggested a Th1-type interleukin-expression pattern. We observed significantly increased Il-2, which acts as a T-cell growth factor as well as elevated Il23 expression, which promotes Th1/Th17 cell differentiation [28, 29]. Both Il-4 and Il-5 mRNA tended to be downregulated, but no significant differences were observed (Fig. 2a). Flow cytometric analysis of splenocytes revealed expansion of CD3+CD4+ and CD3+CD8+ T-cells. The percentage of activated CD3+CD8+CD44+ cell subset was also significantly increased in Gdf15−/−lpr/lpr animals (Fig. 2b). Despite that, lower frequencies of CD4+FoxP3+ regulatory T cells were observed (Fig. 2c). This was associated, with upregulation of B-cell-differentiating transcripts, i.e., Prdm-1 and April and increased frequencies of B220+ MHCII+ CD19+ B cells in spleens of Gdf15−/−lpr/lpr mice (Fig. 2d–f). Despite increased Flt3l levels, we found reduced CD11c+ dendritic cells frequency and similar levels of F4/80+ myeloid cells in spleens of Gdf15−/−lpr/lpr compared with C57BL/6lpr/lpr mice (Fig. 2g, h). Absolute cell counts of these cells can be retrieved from online supplementary Figure 1. Thus, GDF15 suppresses the proliferation and activation of T- and B- cells in the context of murine systemic autoimmunity.
GDF15 Suppresses Type I Interferon Signature, Autoantibody Production, and Renal Damage in lpr Mice
Next, we assessed female Gdf15−/−lpr/lpr mice for two hallmarks of lupus-like disease namely, type I interferon production and humoral autoimmunity. Whole spleen mRNA analysis revealed upregulation of type I and III interferon-dependent signaling pathways, i.e., increased Ifit-1, Ifit-3, Zbp1, Ifnλ2, and Ifnλ3 expression (Fig. 3a, b). Moreover, classical Th1-cytokines, i.e., IFNγ, IL-12, and TNFα were increased in Gdf15−/−lpr/lpr mice (Fig. 3c). Despite lower frequencies of B220+ CD11c+ CD137+ “plasmacytoid dendritic” cells, GDF15-deficiency promoted increased serum levels of IFNα in autoimmune mice (Fig. 3d, e). We then assessed the splenic expression of plasma cell maturation repressor BLIMP-1 which drives the terminal differentiation of B cells to plasma cells . As expected we observed increased numbers of B220+ and BLIMP-1+ cells in Gdf15−/−lpr/lpr mice (Fig. 3f). Moreover, we detected more CD138+ kappa-light-chain+ plasma cells in spleens of Gdf15−/−lpr/lpr mice (Fig. 3g). This was associated with increased total IgM and IgG-serum levels (Fig. 3h; online suppl. Fig. 2) and higher levels of anti-ds-DNA- and anti-histone-autoantibodies in Gdf15−/−lpr/lpr animals, whereas anti-RNA/smith-directed auto-antibodies did not significantly differ (Fig. 3i).
These serologic features of SLE were accompanied by a mild renal phenotype in female Gdf15−/−lpr/lpr mice evidenced by increased serum creatinine levels, increased proteinuria and increased urinary albumin-creatinin-ratios (Fig. 4a). CD45-and MAC2-stainings revealed increased glomerular infiltration of leukocytes and predominantly macrophages in glomerular compartments (Fig. 4b). Despite similar complement activation, slightly increased glomerular IgG deposition and reduced counts of podocytes were found (Fig. 4c).
Taken together GDF15 accelerates the progression of autoimmune disease. It promotes enhanced interferon production, plasma cell maturation, autoantibody secretion, and results in aggravated nephritis without crescent formation.
GDF15 Negatively Regulates TLR-9 Dependent Proliferation and TLR-7 Dependent Serologic Features of Autoimmunity
Our data show that GDF15 deficiency alone was not sufficient to trigger autoimmunity at the age of 6 months. Yet, what could be the mechanism of how GDF15-deficiency aggravates autoimmunity in the lupus-prone C57BL/6lpr/lpr strain? Gdf15−/−lpr/lpr mice showed increased splenocyte cell death and increased cell-free DNA (Fig. 1h–j). Moreover, the increased production of serum inflammatory mediators like TNFα, IL-12, IFNγ, or IFNα (Fig. 3c, d) might be a possible mechanism. Still, no differences were detected in Gdf15−/− C57BL/6 versus C57BL/6 mice in these experiments (Fig. 3c, d). Comparable systemic corticosterone and aldosterone levels argue against a central immunomodulatory role of the GDF15-GFRAL-axis  (Fig. 5a).
Since endosomal TLRs, i.e., TLR-7 and -9 are involved in lupus pathogenesis, i.e., lymphocyte proliferation, as well as type-I interferon and anti-dsDNA-autoantibody production we tested splenic Tlr7 and Tlr9 expression and found the latter mildly increased in GDF15-deficient mice (Fig. 5b). Next, we stimulated macrophages of Gdf15-intact genotypes with the TLR-9 ligand CpG-DNA and assessed GDF15 levels from supernatants. Here, both wild-type C57BL/6 and lupus-prone C57BL/6lpr/lpr mice produced a significantly increased quantity of GDF15 (Fig. 5c). Splenocytes of GDF15-deficient mice showed enhanced proliferative responses after stimulation with CpG-DNA. This Gdf15 dependent effect was apparent in C57BL/6 and lupus-prone C57BL/6lpr/lpr background (Fig. 5d). Further, stimulation of BMDMs with the TLR-9 and -7 ligands, CpG and Imiquimod resulted in increased expression of interferon dependent genes, i.e., type I interferon negative regulator Ifi204, Ifit-1, and Ifit-3 in GDF15 deficient mice. This effect was observed for Ifi204 after CpG and for Ifit-1 and Ifit-3 after Imiquimod exposure in both lpr- and non-lpr C57BL/6 strains (Fig. 5e). Consequently, Gdf15 and Gdf15−/−lpr/lpr macrophages produced significantly more TNFα upon CpG stimulation. Stimulation of IFNγ-activated macrophages with CpG resulted in significantly increased IL-12 production (Fig. 5f). Of note, we performed similar experiments with cDCs and pDCs and did not find significant differences in response to CpG and Imiquimod stimulation with respect to interferon-dependent gene expression (online suppl. Fig. 3). Thus, GDF15 suppresses TLR-9 and -7 signaling-dependent pro-inflammatory and interferon-dependent gene expression in macrophages as well as TLR-9-dependent lymphocyte proliferation. As depicted in Figure 5f, combined stimulation of BMDMs with IFNγ+ CpG-DNA mimicking a lupus-like micro milieu resulted in significantly increased IL-12 and TNFα release in Gdf15−/− and Gdf15−/−lpr/lpr versus controls. Still, the induction of the interferon-dependent genes Ifit-1 and Ifit-3 was only enhanced upon activation of TLR-7 in Gdf15−/− animals versus controls (Fig. 5e). Taken together, these in vitro data demonstrate a negative regulatory role for GDF15 on TLR-9 and TLR-7 signaling.
GDF15 Expression Is Downregulated in Human LN
Our data highlight the role of Gdf15 in controlling serologic activity in murine SLE but does not investigate the relevance of GDF15 in human disease. Kidney involvement is the most common visceral organ manifestation in human SLE . Therefore, we examined glomerular and tubular kidney tissue from patients with LN for abnormal expression of GDF15 and interferon-dependent genes (i.e., IFIT-1 and IFIT-3). Biopsy material from living kidney donors (LD) served as reference values. Kidney specimens affected from MCD or HTN served as “low-inflammatory” controls. In glomerular compartments, GDF15 expression was downregulated in LN, as well as HTN and MCD specimens. Still, only in LN, we have found a concomitant strong upregulation of interferon-dependent gene- and MKI67- expression (Fig. 6a). In LN, whereas glomerulonephritis is regarded as a manifestation of systemic autoimmunity, tubulointerstitial inflammation is thought to reflect a prognostically relevant local inflammation-amplification-loop . Interestingly, in tubular compartments, we observed the same pattern of downregulated GDF15, yet upregulated type I interferon-dependent gene expression (Fig. 6b). To assess whether GDF15 mRNA levels were related to changes in interferon-dependent genes (i.e., IFIT-1 and IFIT-3) expression, we performed Spearman’s correlation analysis among LN patients and healthy controls. We found a significant negative correlation between GDF15 and tested interferon-dependent genes in glomerular (rho = −0.30 and −0.36, respectively, p < 0.01) compartments. This was even more pronounced in tubular compartments of LN kidneys (rho = −0.50 and −0.53; p < 0.001).
We had hypothesized a negative-regulatory role for GDF15 in the pathogenesis of systemic autoimmunity, i.e., murine lupus-like disease. The presented data support this concept as GDF15 acted as a suppressor of lymphoproliferation, B- and T-cell expansion, autoreactive plasma-cell-maturation, and consequently anti-DNA-directed autoantibody generation in the C57BL/6lpr/lpr model. Moreover, we found, that GDF15 suppresses type I interferon signature in lupus-prone mice. Yet, consistent with our previous data on negative regulators of TLR-signaling [32-34], an additional factor creating an autoimmune-convenient environment was required before this immunoregulatory role of Gdf15 came into play.
Both, type I interferon and anti-dsDNA-autoantibody production, are hallmarks of systemic lupus or lupus-like disease. The type I interferon signature has been causally linked to SLE pathogenesis via (a) genetic gain-of-function variants-associated risk for disease, (b) induction of SLE-like phenotypes in interferon treated cancer patients, and (c) correlation of type I interferons with antinuclear antibodies (ANA)-titers, disease activity, and nephritis in lupus patients . Lastly, anifrolumab, an inhibitory human IgG1κ monoclonal antibody to type I interferon receptor subunit 1, showed promising results in the recent TULIP-2 trial performed in patients with active SLE . Gdf15−/− in lupus-prone C57BL/6lpr/lpr mice significantly enhanced the serologic IFNα and IFNγ production and enforced an inflammatory systemic cytokine profile of increased systemic TNFα and IL-12p70 serum levels . These data are consistent with previously published in vivo data on diabetic mice, where GDF15 was shown to inhibit pro-inflammatory macrophage polarization .
Our data highlight TLR-7/-9 hyperresponsiveness as possible mechanisms of the exaggerated type I signature in GDF15-deficient C57BL/6lpr/lpr mice. Although, plasmacytoid dendritic cells were considered the classical interferon producers in SLE, in the meantime macrophages have also been implicated in this process [38-41]. In Gdf15−/− C57BL/6lpr/lpr mice, the frequency of B220+ CD11c+ CD137+ dendritic cells was not enhanced in spleens and ex vivo TLR-7/-9 hyperresponsiveness was detected in BMDMs but not BMDCs, consistent with an important role of macrophages in type I IFN production.
Apart from exaggerated serologic autoimmune activity, GDF15- deficiency was associated with a mild nephritic phenotype in mice as evidenced by increased glomerular IgG deposits, mild hypercellularity, reduced podocytes numbers, and increased albumin creatinine ratios. However, we did not observe full-blown crescentic LN. One might argue that 6 months of observation might have been too short to observe full-blown nephritis. Yet, similar experiments by our group investigating other negative regulators of innate immunity, e.g., SIGIRR or the NLRP3 inflammasome components revealed stronger histologic evidence of renal injury or bronchiolitis/pneumonitis at the same age [32, 42].
In acute murine anti-GBM nephritis in C57BL/6 mice, GDF15-deficiency promoted CXCL10/CXCR3-dependent T-cell infiltration and crescent formation . In addition, Liu et al.  recently demonstrated that increased GDF15 level in tubular epithelial cells upon ischemia-reperfusion injury prevented post-ischemia inflammation and injury. Although GDF15 has a similar renal protective effect in all these models, its effect size varies between acute versus chronic inflammation.
The question remains of how GDF15 flips the immunoregulatory switch in murine lupus-like autoimmunity. Up to recently, only the CNS receptor GFRAL had been a confirmed interaction partner of GDF15 . Luan et al.  recently proposed a model in which GDF15 acts as a central mediator of inflammatory tissue tolerance in sepsis via a central regulatory mechanism controlling hepatic lipid export in a GDF15-GFRAL-sympathetic-axis-dependent manner. Yet, C57BL/6 and C57BL/6lpr/lpr mice showed comparable systemic levels of GDF15 at the age of 6 months when autoantibodies and lymphoproliferation were already established and systemic corticoid hormone levels were comparable amongst Gdf15−/− and control mice. Therefore, a central regulator mechanism seems unlikely in our model.
Instead, it was recently shown that GDF15 suppresses anti-(HCC)-tumor immunity via CD48 engagement on Tregs . In fact, Gdf15−/− C57BL/6lpr/lpr mice displayed reduced peripheral Treg counts. Reduced Tregs however are a frequent finding in an active murine and human SLE . Scurfy mice, which display a missense mutation in the transcription factor Foxp3 and are deficient in Tregs, develop early-onset hyperinflammation, autoantibodies, and lymphoproliferation . Thus, reduced regulatory T-cell activity, which we were unable to assess in vivo, could explain some aspects of the observed phenotype but falls short of explaining the prominent type I interferon/interferon-alpha production in our mice .
Instead, we demonstrate a direct immunosuppressive effect of Gdf15 on macrophage TLR7 and TLR-9-signaling in C57BL/6 and C57BL/6lpr/lpr mice in vitro. Similarly, others have demonstrated that GDF-15 suppresses macrophage inflammatory cytokine release and T-cell activation in alcohol-induced chronic liver injury in mice . While most phenotypic alterations in Gdf15−/− C57BL/6lpr/lpr mice, e.g., increased cfDNA, type I interferon- and anti-dsDNA-autoantibody production were absent in Gdf15−/− C57BL/6 mice, TLR-9-dependent-splenocyte proliferation, TLR-9-dependent-macrophage-cytokine secretion, and TLR7 dependent Ifit-1 and -3 expression were already apparent in Gdf15−/− C57BL/6 mice. In accordance, type I interferon signature and anti-dsDNA autoantibody production is closely tied to aberrant signaling of innate DNA-sensors in the early pathogenesis of SLE [47, 48]. Immune complexes are known ligands of TLR-7 and -9 and inducers of type I interferon in SLE macrophages .
Targeting TLR-7 with monoclonal antibodies was shown to reduce circulating monocytes and ameliorate LN in NZBWF1 mice . Similarly, overexpression of TLR-7 aggravates type I interferon production, anti-RNA autoantibody secretion and nephritis in mice . TLR-9-signaling on the other hand is relevant for anti-DNA autoantibody production but also protects against LN in lpr mice. We therefore consider both, increased TLR-9- and TLR-7-signaling, to be relevant for the immunologic features of Gdf15−/− C57BL/6lpr/lpr mice: autoantibodies and type I interferons (TLR-7 and TLR-9) versus LN (TLR-7) . Clearly, this remains speculative, since the combined deletion of TLR-7/9 and GDF15 would be necessary to prove these considerations.
Our study has several limitations. Since we were unable to observe full-blown nephritis and mortality of Gdf15−/− C57BL/6lpr/lpr before the experiment was terminated. In addition, in vivo relevance of the proposed GDF15/TLR-inhibitory mechanism needs to be further investigated and requires a strategy involving combined knock out of multiple genes encoding Tlr7, Tlr9, and Gdf15.
Taken together, we demonstrate that GDF15 displays a disease-modifying effect in lupus-like autoimmunity. Thus, GDF15 suppresses lymphoproliferation, type I interferon-, and anti-DNA-autoantibody production in autoimmune-prone mice. Human data further indicates its involvement in human LN.
We thank all participating centers of the European Renal cDNA Bank – Kröner-Fresenius biopsy bank (ERCB-KFB) and their patients for their cooperation. Active members at the time of the study see .
Statement of Ethics
This study was carried out following the principles of the Directive 2010/63/EU on the Protection of Animals Used for Scientific Purpose and with approval by the local government authorities (ROB Az 55.2-1-54-2531-11-10). The study follows the ARRIVE guidelines. Human kidney biopsy specimens and Affymetrix microarray expression data were procured within the framework of the European Renal cDNA Bank-Kröner-Fresenius Biopsy Bank. Biopsies were obtained from patients after informed consent and with the approval of the local Ethics Committees (the Ethics Committee Reference Number #250-16 [Ethics Committee of medical faculty at LMU Munich]). Written informed consent was obtained from the patients.
Conflict of Interest Statement
The authors have no conflicts of interest.
The study was funded by the Deutsche Forschungsgemeinschaft (LE2621/6-1) and National Science Center, Poland Grants 2016/23/B/NZ6/00086 to ML. The ERCB-KFB was supported by the Else Kröner-Fresenius Foundation.
Georg Lorenz and Maciej Lech performed experiments and data analysis, designed the study, and wrote the manuscript. Andrea Ribeiro, Ekatharina von Rauchhaupt, Vivian Würf, Twinkle Vohra, and Maja Lindenmeyer performed experiments and data analysis. Hans-Joachim Anders, Christoph Schmaderer, Clemens D. Cohen, and Maja Lindenmeyer carefully read and improved the manuscript and were involved in data interpretation.
Data Availability Statement
Published gene expression profiles (Affymetrix GeneChip Human Genome U133A and U133 Plus2.0 Arrays; GSE99340, GSE32591, GSE35489, GSE37463) used in this study.