Autoimmune Kidney Disease and Impaired Engulfment of Apoptotic Cells in Mice with Macrophage Peroxisome Proliferator-Activated Receptor γ or Retinoid X Receptor α Deficiency

Systemic lupus erythematosus (SLE) is an autoimmune disorder that can present with a renal involvement that leads to SLE nephritis (1). SLE nephritis develops as a consequence of autoreactive B cell activation, production and intrarenal deposition of autoantibodies and complement components, macrophage activation, and the generation of proinflammatory cytokines. SLE nephritis disrupts renal function, contributes to the development of peripheral vascular disease, and increases the risk of cardiovascular mortality (2).

The disease mechanism of SLE was shown recently to involve impaired clearance of apoptotic cells, which has led to a conceptual change in our understanding of the disease (3). The nuclear DNA fragments and nuclear proteins of dying cells are self-Ags, and they are important triggers of autoreactive Ig production (4). The safe removal of apoptotic cell debris is a dedicated function of phagocytic cells, particularly tissue-resident macrophages (5, 6). Signals from apoptotic cells direct the activation of macrophages toward a so-called regulatory or deactivated phenotype, thereby dampening inflammation (7–9). Macrophages reprogrammed by apoptotic cells control and deactivate immunity and inflammation by triggering the release of anti-inflammatory cytokines, such as TGF-β and IL-10, and inhibiting the production of proinflammatory cytokines. In pathologies in which apoptotic cell clearance is disrupted, as in SLE, the absence of this suppressive effect on inflammatory cytokine production by macrophages could aggravate chronic inflammation (10). To correctly engulf apoptotic cells, macrophages express a large set of cell surface receptors, including scavenger receptors (e.g., CD36, CD14), integrins (e.g., ITGA5, ITGB3), Ig supergene family receptors (e.g., FcγRs), receptors for complement factors (e.g., LRP1/Calr, the low-density lipoprotein receptor-related protein 1/calreticulin complex), complement receptor 3 (CR3, also known as ITGM), complement receptor 4 (CR4, also known as ITGX), and Tyro3/Axl/Mertk (TAM) family tyrosine kinase cell surface receptor. Apoptotic cell removal is enhanced by bridging molecules (opsonins), which interact with macrophage receptors and apoptotic cell molecules. Opsonins include components of the complement system (C1q, C3b, C4b, and iC3b), immunoglobulins (IgG and IgM), MFG-E8, THBS1, and GAS6. Altered expression of some cell-surface receptors and opsonins has been proposed to influence the development of SLE (11–13).

Ligands of peroxisome proliferator activated receptor γ (PPARγ) modulate macrophage function and regulate the expression of a range of inflammatory genes, making them potential drugs for the treatment of inflammation and immune responses (14). Activators of PPARγ include lipid metabolites and synthetic thiazolidinediones such as rosiglitazone (RSG) (15). PPARγ belongs to the nuclear receptor superfamily of ligand-dependent transcription factors and affects various cellular processes at the level of gene expression regulation (16, 17). PPARγ controls macrophage polarization and inhibits expression of inflammatory genes (14); therefore, it is a key regulator of macrophage-mediated immune responses (15, 18). PPARγ forms a permissive heterodimer with retinoid X receptors (RXRs). These nuclear receptors, encoded by three distinct genes (Rxra, Rxrb, Rxrg), are heterodimer-forming partners of other nuclear receptors (19). The most important RXR isoform in macrophages is RXRα (20), which is activated by its natural ligand 9-cis-retinoic acid (9cRA) and by the synthetic ligand LG268 (21). Permissive PPARγ/RXRα heterodimers can be activated by ligands of either receptor and have additive or synergistic downstream effects, suggesting that PPARγ might modulate macrophage functions through heterodimerization with RXRα (16, 19, 22, 23). However, the immunomodulatory role of PPARγ/RXRα heterodimers in macrophages is still undefined.

In this study, we show that macrophage-specific deletion of PPARγ or RXRα leads to glomerular damage and an elevated production of autoantibodies to nuclear Ags, both hallmarks of SLE. Mice lacking macrophage PPARγ or RXRα show impaired engulfment of apoptotic cells and a deficient expression of a gene network implicated in phagocytosis. These results indicate that PPARγ and RXRα signaling in macrophages is essential for the efficient clearance of apoptotic cells and the maintenance of self-tolerance, an important element of tissue homeostasis and a mechanism disrupted in SLE. The impaired uptake of apoptotic cells by macrophages is an initiator of autoimmunity, making macrophage PPARγ and RXRα candidate targets for new SLE medication strategies (1).

Mice carrying floxed alleles of PPARγ (PPARγ^f/f) or RXRα (RXRα^f/f) were generated as described previously (18, 24). We used a conditional knockout (KO) mouse model with myeloid-specific PPARγ ablation to generate mice lacking macrophage PPARγ or RXRα (20). Mice bearing the lox-P-targeted PPARγ (PPARγ^f/f) or RXRα (RXRα^f/f) allele and the lysozyme-M Cre (LysMCre) recombinase transgene were crossed to generate offspring with macrophage-specific loss of the PPARγ or RXRα gene. We refer to LysMCre-negative PPARγ^f/f and RXRα^f/f mice as wild type (WT), and their PPARγ^f/fLysMCre⁺ or RXRα^f/fLysMCre⁺ littermates as knockout animals (PPARγ-KO and RXRα-KO). All mice were on a 129/Sv and C57BL/6 background. Female mice aged 4–6 or 14–16 mo were used for experiments. Mice were genotyped by PCR (18). Experiments were performed according to local ethical guidelines and were approved by the Animal Subjects Committee of the Instituto Carlos III (Madrid, Spain) in accordance with EU Directive 86/609/EEC. Mice were killed by CO₂ inhalation.

To characterize glomerular angioarchitecture, kidneys were perfused with water resistant ink through the renal arteries and immediately fixed in 10% paraformaldehyde (PFA) diluted in PBS (0.1 M, pH 7.4) for 24 h. Samples were prepared for optical projection tomography (OPT) by a processing protocol that provides high tissue transparency to allow scanning of arterial casts (25). OPT was performed with a Bioptonics Microscopy OPT Scanner 3001 (Bioptonics, Edinburgh, Scotland).

For histologic analysis, kidneys were dissected, weighed, and fixed in 4% PFA for 24 h. Tissues were dehydrated, cleared in xylene, and finally embedded in Paraplast embedding medium (Structure Probe, West Chester, PA). Microtome sections (5 μm) were stained with Harris H&E or periodic acid-Schiff (PAS) according to the manufacturer’s protocol (Sigma-Aldrich, St, Louis, MO). Glomerular diameter and the number of glomerular cell nuclei were determined by histomorphometric analysis of digital images taken at ×400 magnification according to a previously described method (26). Apoptotic cells were labeled with either the TUNEL Apoptag Plus Peroxidase Kit (Chemicon, Temecula, CA) or the In Situ Cell Death Detection Kit (Roche, Basel, Switzerland), or they were immunostained for cleaved caspase 3 (Cell Signaling Technology, Beverly, MA). The apoptotic index, the percentage of TUNEL positive nuclei in glomeruli, was calculated from counts of more than 60 glomeruli per sample. For silver impregnation of basement membranes, we used the PAS-methenamine-silver method (IHC Worldwide). Immunostainings for mouse F4/80 (Abcam, Cambridge, U.K.), CD68 (Santa Cruz Biotechnology, Santa Cruz, CA), vascular endothelial growth factor-A (VEGF-A; Abcam), TNFα (Santa Cruz Biotechnology), inducible NO synthase (iNOS; Santa Cruz Biotechnology), and IL-10 and IL1β (Abcam) were performed on vibratome sections following the manufacturer’s protocols. For visualization, we used the Vectastain ABC Elite Universal Kit (Vector Laboratories, Burlingame, CA). Distribution of mouse IgG and IgM was determined by Fab-specific anti-mouse IgG and μ-specific anti-mouse IgM (27) (FITC conjugated, developed in goat; Sigma-Aldrich).

For analysis of cell ultrastructure, samples were cut from the kidney cortex and immersed immediately in a mixture of PFA and glutaraldehyde (4/1). Samples were dehydrated and embedded in Durcupan resin (Sigma-Aldrich). Semithin sections were osmificated and uranyl acetate contrasted, and finally ultrathin sections were mounted on grids and examined with a JEOL JEM 1010 transmission electron microscope. Images were taken with a Sony digital camera.

Animals were kept individually in metabolic chambers (Tecniplast, Italy) and urine samples were pooled over 24 h. Total urinary albumin was determined by latex turbidimetry (Turbilatex; Spinreact; Sant Esteve d'en Bas, Spain). Serum creatinine was measured from tail vein blood samples by an enzymatic endpoint colorimetric method (Spinreact). Measurements were recorded with a Spinlab 100 clinical chemistry analyzer (Spinreact, Girona, Spain). Urinary proteins were separated by continuous 12% SDS-PAGE. Serum VEGF-A was measured with a mouse VEGF-A ELISA kit (Bender MedSystems, Vienna, Austria), and serum titers of autoreactive Abs (anti-ssDNA, anti-dsDNA and anti-nuclear Ab [ANA]) were determined with an ELISA kit (αDiagnostics, San Antonio, TX).

Cell suspensions were prepared from spleen, axillary and neck lymph nodes, and freshly harvested femur bone marrow from WT, PPARγ-KO, and RXRα KO mice. The cells were analyzed by FACS. The following labeled Abs were used to detect different leukocyte subpopulations: FITC-labeled anti–c-Kit and anti-IgM, PE-labeled anti-Ly6G and anti-CD4, Alexa Fluor 647-labeled anti-Cd11b and B220, Alexa Fluor 488-labeled Ly6C and CD3, PE-Cy7–labeled anti-F4/80, Pacific Blue-labeled anti-CD45, allophycocyanin-labeled anti–SCA-1, PerCP-labeled anti-CD8, and biotin-conjugated Lin. Cells were analyzed with a FACSCalibur flow cytometer and CellQuest software (BD Biosciences, San Jose, CA).

Apoptotic thymocytes (ATs) were generated and fluorescently labeled as described (28). Thymocytes were harvested from 4-wk-old C57BL/6 mice. Cells were treated with 1 μM dexamethasone (Sigma-Aldrich) for 6 h at 37°C and were then FITC-labeled with CellTracker green (Invitrogen, Carlsbad, CA). Efficient induction of apoptosis was confirmed by flow cytometry using the Annexin-FITC Apoptosis Detection Kit (BD Pharmingen, San Diego, CA). Dexamethasone treatment yielded an AT population with ∼70% apoptotic cells (Supplemental Fig. 2A).

Fluorescently labeled ATs (2 × 10⁶) were injected i.p. into WT, RXRα-KO, or PPARγ-KO mice. Animals were sacrificed after 6 or 14 h, peritoneal cells were retrieved by lavage, and free ATs were examined with a Leica AM TIRF MC wide field fluorescent-differential interference contrast microscope (Leica Microsystems, Deerfield, IL) and counted with a hemocytometer. For ultrastructural analysis, macrophages retrieved by lavage were pooled in 5% agar gel, fixed in a mixture of PFA and glutaraldehyde (4/1), and processed for electron microscopy as described above.

The phagocytosis assay was performed as described previously (29). Peritoneal exudates were collected by lavage from mice treated with 3% thioglycollate 3 d before cell isolation. The macrophage cell population was obtained after negative selection with magnetic bead-conjugated rat anti-mouse Ly-6G Ig (eBioscience, San Diego, CA). Cells were plated on coverslips in 24-well culture plates in RPMI 1640 medium supplemented with 2% FBS (Life Technologies, Rockville, MD) and penicillin-streptomycin. To test the effects of PPARγ and RXRα ligands, macrophages were treated for 18 h with 1μM RSG (Cayman Chemical, Ann Arbor, MI) or 100 nM 9cRA (Sigma-Aldrich). Before the phagocytosis assay, macrophages were starved for 2 h in RPMI 1640 containing 20 mM Hepes, l-glutamine, and penicillin-streptomycin. ATs were added to macrophages at a 5:1 ratio and incubated for 1 h. Wells were washed five times with PBS and cells were fixed with 1% PFA. Coverslips were mounted with Prolong Gold antifade reagent containing DAPI (Invitrogen; Molecular Probes, Eugene, OR). RBCs from sheep blood (Cappel, ICN Pharmaceuticals, London, U.K.) were suspended in PBS containing 20 mM glucose and mixed 1:10 with anti-sheep RBC IgG (Cappel) or 1:7 with anti-sheep RBC IgM (Accurate Chemical, Westbury, NY) and incubated for 20 min at room temperature. IgM-opsonized RBCs were further incubated for 20 min at 37°C with C5-deficient human serum (Sigma-Aldrich). Opsonized RBCs were suspended in culture medium and added to macrophages at a 5:1 ratio. After 30 min incubation, cells were fixed in 4% PFA. After several washes in PBS, opsonized RBCs were stained with FITC-conjugated goat anti-rabbit IgG (Invitrogen). Macrophages were labeled with Alexa Fluor 488 phalloidin (Invitrogen). In control experiments, macrophage cultures were incubated for 30 min with FITC-labeled dextran microspheres (Sigma-Aldrich).

The proportion of macrophages containing ingested particles was determined by manually counting >100 macrophages per sample. The phagocytosis index was calculated according to the formula: (number of macrophages containing one ingested particle + number of macrophages containing two ingested particles × 2 + number of macrophages containing three ingested particles × 3 + number of macrophages containing four or more ingested particles × 4) ÷ total number of counted macrophages.

For dynamic imaging, macrophages were cultured in glass-bottom culture dishes (MatTek, Ashland, MA) as described above, and AT uptake was monitored in real time with a Leica AM TIRF microscope in fluorescent widefield-brightfield mode, using a ×40/1.25 oil-immersion objective. Images were acquired with an Andor iXon EMCCD chip.

Peritoneal macrophages were harvested as described previously (18). For ligand assays, macrophages were incubated with 100 nM RSG, 100 nM 9cRA, or 100 nM LG268 (a gift from Reid Bissonette, Ligand Pharmaceuticals, La Jolla, CA) for 24 h. For AT assays, ATs were added to macrophages (5:1) for 30 min or 24 h. When needed, ATs where removed by gentle washing, and macrophages were incubated with 100 ng/ml LPS for an additional 5 h for gene expression analysis or 18 h for cytokine secretion studies. Cytokine secretion was quantified in culture supernatants using commercially available ELISA kits (BD Biosciences). Total macrophage RNA was isolated using Trizol (Sigma-Aldrich). Transcripts were quantified in a two-step RT-quantitative PCR process. First-strand cDNA was synthesized from 1 μg total RNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA) with random hexamers. Samples were run in 20 μl reactions using the GeneAmp PCR System 9700 (Applied Biosystems). Quantitative PCR reactions and analyses were then performed using SYBR Green oligonucleotides and the ABI PRISM 7900 FAST Sequence Detection System (Applied Biosystems). The analysis was customized, with two replicates per target gene and two parallel analyses for each biologic sample. Expression levels of target genes were normalized to a standard housekeeping gene (36b4) as an internal control. Gene expression values were expressed as relative mRNA level (fold change) compared with untreated controls, based on the standard curve method as described by the manufacturer (Invitrogen). Primer sequences can be provided upon request. Affimetrix U74Av2 microarrays were analyzed as described previously (30). Expression data were visualized using Treeview (Taxonomy and Systematics, Glasgow, U.K.). The microarray data sets have been submitted to the GEO database under accession number GSE24292 (http://www.ncbi.nlm.nih.gov/geo/).

The generation and analysis of the ChIP-seq data for PPARγ binding and histone H3 lysine 9 acetylation (H3K9Ace) enrichment in mouse macrophages have been described elsewhere (31). Chromatin immunoprecipitations for PPARγ (sc-7196; Santa Cruz Biotechnology) and H3K9Ace (ab4441; Abcam) were performed using primary WT thioglycollate-elicited macrophages, followed by amplification of the precipitated DNA and sequencing of 36-bp reads using the Solexa Analysis Pipeline. The reads were mapped to the mouse genome (mm8) using ELAND. Peak calling for PPARγ was performed using GLITR (32) and the following thresholds were applied to obtain high-confidence binding regions: false discovery rate (FDR) ≤0.5%; fold change over input ≥5; number of reads at the peak summit ≥9. In this study, the FDR threshold was also lowered to 5%, and the additional requirements of fold change and number of reads at the peak summit were eliminated to obtain a less stringent dataset. H3K9Ace data were analyzed using Significance Tester for Accumulation of Reads (http://www.cbil.upenn.edu/STAR) using a threshold of FDR 5% (33). Areas of input bias, based on peak enrichment in input sequencing runs, were masked in all datasets.

Acquisition data are summarized in Supplemental Table 1. Images were acquired with a Nikon BF/Fluo microscope or a Leica AM DIC/TIRF microcope, using an ×40/1.25 oil-immersion objective and an Andor iXon EMCCD chip. We used NIS Elements acquisition software.

All data are means ± SEM. Data from histomorphometric analysis, clinical chemistry measurements, and phagocytosis assays were compared by two-tailed paired Student t tests. Quantitative PCR data were analyzed by unpaired Student t test for comparison of WT and KO samples and by paired Student t test for comparison of untreated and treated WT cells. The normal distribution of data were checked by the Kolmogorov-Smirnov test, with α = 0.05.

Conditional KO mice lacking macrophage expression of PPARγ (18) or RXRα (24) were generated by crossing mice bearing the lox-P-targeted PPARγ (PPARγ^f/f) or RXRα (RXRα^f/f) alleles with mice bearing the lysozyme-M Cre (LysMCre) recombinase transgene. We refer to LysM Cre-negative PPARγ^f/f and RXRα^f/f mice as WT, and their PPARγ^f/fLysMCre⁺ or RXRα^f/fLysMCre⁺ littermates as KO animals (PPARγ-KO and RXRα-KO).

Female mice lacking macrophage expression of PPARγ or RXRα developed severe nephropathy at 4–6 mo old. An early sign of nephritis is enlargement of the renal glomeruli (12), so we analyzed the kidney angioarchitecture by OPT. OPT imaging of renal arteriolar casts revealed enlarged glomeruli in focal regions within the kidneys of PPARγ-KO and RXRα-KO animals (Fig. 1A). Kidney hypertrophy, a consequence of glomerulomegaly, caused a significant increase in kidney weight in PPARγ-KO and RXRα-KO mice compared with WT littermates (Supplemental Fig. 1A).

Histologic analysis revealed mesangial hypercellularity, mesangial matrix expansion, and basement membrane thickening in PPARγ-KO and RXRα-KO kidneys, and these histopathologic alterations were more evident in older mice (Fig. 1B). The increases in glomerular size and glomerular cell number in the KO mice were confirmed by histomorphometric analysis (Fig. 1C), and glomerular basement membrane thickening was also visible by silver impregnation (Fig. 1D). These histopathologic features of PPARγ-KO and RXRα-KO mice are hallmarks of glomerulonephritis (GN) (34). Moreover, we also detected excretion of albumin (Fig. 1C) and high m.w. proteins in the urine of PPARγ-KO and RXRα-KO mice (Supplemental Fig. 1B), as well as elevated plasma creatinine levels (Fig. 1C), all of which are possible consequences of GN (35, 36).

Transmission electron microscopy (TEM) of nephrons demonstrated that glomerular basement membranes were thickened by 38 ± 6% in PPARγ-KO mice and 35 ± 4% in RXRα-KO mice compared with their WT littermates. TEM also detected coarsening and fusion of podocyte foot processes in PPARγ-KO and RXRα-KO samples (Fig. 1D). These findings are consistent with the ultrastructural changes reported in mice with proteinuria (35). Consistently, cell destruction in the glomeruli in PPARγ-KO and RXRα-KO mice was indicated by the accumulation of electrolucent material in podocyte foot processes (Fig. 1D).

To further characterize the glomerular injury, we examined the localization of inflammatory markers and mouse immunoglobulins by immunohistochemistry. Most glomeruli in the kidneys of PPARγ-KO and RXRα-KO mice contained numerous F4/80-positive inflammatory cells (Fig. 2A). These glomeruli also showed strong immunostaining for VEGF-A in the mesangium surrounding the glomerular capillaries, correlating with elevated serum levels of VEGF-A in PPARγ-KO and RXRα-KO mice (Fig. 2A, 2C). These findings indicate severe glomerular inflammation associated with massive macrophage infiltration (37). In addition, there was a marked deposition of IgG and IgM in the mesangial matrix of PPARγ-KO and RXRα-KO mice (Fig. 2B), and analysis of serum showed that these animals developed autoantibodies against nuclear proteins (ANA) (Fig. 2D), ssDNA (anti-ssDNA; Fig. 2E) and dsDNA (anti-dsDNA; Fig. 2F), suggesting an autoimmune origin of the observed GN (1). To determine whether the GN results from changes in lymphoid and myeloid cell subsets, we performed an immunophenotyping analysis of bone marrow, spleen, and lymph nodes in WT, PPARγ-KO, and RXRα-KO mice. The numbers of different leukocyte populations did not differ between WT, PPARγ-KO, and RXRα-KO littermates (Supplemental Fig. 3), suggesting that the expansion of immune cells is not the cause of the observed GN.

Impaired clearance of apoptotic cells within the kidney glomeruli is thought to contribute to autoimmune GN (12, 38). A significant increase in the number of apoptotic cells in the glomeruli of PPARγ-KO and RXRα-KO mice was detected by TUNEL (Fig. 3A, 3B). Consistently, PPARγ-KO and RXRα-KO glomeruli were abundant in endothelial and mesangial cells containing cytosolic cleaved caspase-3 (Fig. 3A). Immunostaining revealed that cleaved caspase-3 positive apoptotic cells were distributed throughout the spleens of PPARγ-KO and RXRα-KO mice and that a significantly lower proportion of splenic macrophages contained apoptotic cells in these mice (Supplemental Fig. 1C), indicating a generalized defect in apoptotic cell clearance (4, 39).

Accumulation of apoptotic cells in the spleen can be caused by insufficient phagocytosis by marginal zone macrophages and can induce the production of autoantibodies against nuclear DNA fragments by marginal zone B cells (12). To investigate whether macrophage phagocytosis was altered in PPARγ-KO or RXRα-KO mice, we injected their peritoneal cavities with fluorescently labeled ATs. Peritoneal lavage showed that ATs were still abundant 6 or 14 h after injection. In contrast, the peritoneal macrophages of similarly treated WT littermates efficiently scavenged most of the injected ATs (Fig. 3C, 3D; Supplemental Fig. 2B, 2C). Phagocytosis deficit of peritoneal macrophages of PPARγ-KO and RXRα-KO mice was also reflected by their notably lower phagosome content and smaller filopodia compared with WT macrophages (Fig. 3D). These findings suggest a profound in vivo defect in apoptotic cell clearance in PPARγ-KO and RXRα-KO mice.

To understand the molecular mechanism underlying the defective apoptotic cell removal detected in our mouse models, we profiled gene expression in WT and knockout macrophages by microarray analysis. Clustering analyses revealed that KO peritoneal macrophages are deficient for the transcription of a set of genes encoding phagocytosis-associated receptors and opsonins (Supplemental Fig. 4A). Quantitative PCR analysis confirmed that the expression of Cd36, Axl, Mertk, Fcgr1, and the C1q subunits C1qa, C1qb, and C1qc was lower in both PPARγ-KO and RXRα-KO macrophages (Fig. 4A). In PPARγ-KO macrophages, we also detected a significant downregulation of genes encoding other phagocytosis related molecules, including Tgm2, Mfge8, Lrp1, Itgam, Itgax, and C3 (Fig. 4A; Supplemental Fig. 4B). In contrast, PPARγ and RXRα deficiency had no effect on the expression of the macrophage receptors Cd14, Calr, and Itga5 or the opsonins Gas6 and Thbs1 (Supplemental Fig. 4B).

The digestion of apoptotic cells by macrophages leads to the accumulation of cellular components, including cholesterol and fatty acids, that could act as endogenous ligands of PPARγ and RXRα (16). To investigate this possibility, we analyzed the expression of phagocytosis-related genes in WT, PPARγ-KO, and RXRα-KO peritoneal macrophages incubated with ATs. The presence of ATs increased the expression of Cd36, Mertk, Axl, Fcgr1, C1qa, C1qb, C1qc, and Tgm2 in WT macrophages, but not in PPARγ-KO and RXRα-KO macrophages (Fig. 4B). Apoptotic cell components thus enhance the expression of macrophage receptors and opsonins required for further apoptotic cell uptake, via a PPARγ- and RXRα-dependent mechanism.

We next tested whether the expression of these genes can be modulated by pharmacologic activation of PPARγ or RXRα. Peritoneal macrophages from WT, PPARγ-KO, and RXRα-KO mice were cultured for 24 h in the presence of PPARγ or RXRα agonists (RSG for PPARγ, 9cRA and LG268 for RXRα). All three agonists selectively induced the expression of Cd36, Mertk, Axl, Fcgr1, and Tgm2 in WT macrophages but not in PPARγ-KO or RXRα-KO cells (Fig. 4C; Supplemental Fig. 5A). Furthermore, the expression levels of C1qa, C1qb, and C1qc were also upregulated in WT macrophages treated with RXRα ligands (Fig. 4C; Supplemental Fig. 5A). None of these C1q isoforms was regulated by RSG, although RSG strongly upregulated the expression of the PPARγ target genes ATP-binding cassette transporter G1 (Abcg1) and adipose differentiation-related protein (Adrp) in WT macrophages (Supplemental Fig. 4E). PPARγ or RXRα agonists did not significantly induce the expression in WT macrophages of the receptors Cd14, Calr, Itga5, Lrp1, Itgam, and Itgax and the opsonins Gas6, Thbs1, C3, and Mfge8 (Supplemental Fig. 4C, 4D).

In PPARγ-KO macrophages, but not in RXRα-KO cells, the RXRα ligand LG268 moderately elevated the transcription of Cd36, Mertk, Axl, Fcgr1, Tgm2, and the C1q isoforms, reflecting the ability of RXRα to form heterodimers with other partners. In contrast, in RXRα-KO macrophages the action of the PPARγ agonist RSG was severely impaired, with only a mild increase in Cd36 expression detected that was significantly below that seen in WT cells (Supplemental Fig. 5B).

Further confirmation that the phagocytosis genes Cd36, Mertk, Axl, Fcgr1, and Tgm2 are directly regulated by PPARγ was obtained by assessing genomic PPARγ binding in peritoneal macrophages (31) by chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq). In this analysis, Cd36, Axl, Fcgr1, and Tgm2 all had high-confidence PPARγ binding regions within 100 kb of their transcription start sites, using an FDR of 0.5% (Fig. 4D, blue track). Relaxing the threshold to an FDR of 5% revealed that additional binding regions near these genes and also near Mertk (Fig. 4D, red track). Importantly, many of these PPARγ binding regions were found to overlap with statistically significant regions of H3K9Ace (Fig. 4D, black track), a chromatin mark often found at enhancers and promoters of active genes (40, 41). Considering the responsiveness to a PPARγ ligand and reduced expression in PPARγ- and RXRα-deficient macrophages, the data strongly suggest that these genes are likely to be under the direct transcriptional control of PPARγ/RXRα.

Phagocytosis of apoptotic cells by macrophages promotes their differentiation to a regulatory or deactivated phenotype that is closely associated with the diminished production of inflammatory mediators and the induction of self-tolerance (5, 7–9). These macrophages are characterized by the production of the anti-inflammatory cytokines TGF-β and IL-10 and the expression of surface markers as the mannose receptor and the signaling lymphocytic activation molecule (8, 42). We found that engulfment of ATs upregulates the anti-inflammatory cytokine genes Tgfb1 and Il10 and the cellsurface marker genes Mrc1 (encoding mannose receptor) and Slam (encoding signaling lymphocytic activation molecule) in WT macrophages. These genes were not affected or were affected to a lesser extent in PPARγ-KO and RXRα-KO macrophages (Fig. 5A). Accordingly, in PPARγ-KO and RXRα-KO mice, splenic macrophages showed a strongly reduced immunostaining for IL-10 (Supplemental Fig. 1D).

To test the possible involvement of PPARγ and RXRα in the regulation of proinflammatory cytokine production, we analyzed inflammatory gene expression in WT, PPARγ-KO, and RXRα-KO macrophages stimulated with LPS in the presence or absence of ATs (Fig. 5B). LPS upregulated the expression of genes encoding IL-12B (Il12b), IL-1β (Il1b), iNOS (Nos2, iNOS), and TNF-α (Tnf). The presence of ATs significantly reduced the transcript levels of these proinflammatory markers in WT macrophages, but had no effect on PPARγ-KO and RXRα-KO macrophages (Fig. 5B). Cytokine secretion was also monitored by ELISA. As expected, the presence of ATs suppressed the release of proinflammatory cytokines by WT macrophages, but had little effect on PPARγ-KO and RXRα-KO macrophages (Supplemental Fig. 6).

Consistently, the kidney glomeruli of PPARγ-KO and RXRα-KO mice were infiltrated with cells expressing proinflammatory markers, such as IL-1β, iNOS, and TNF-α (Fig. 5C). These data indicate that the observed GN in mice lacking macrophage expression of PPARγ or RXRα might be a consequence of impaired apoptotic cell removal and enhanced proinflammatory macrophage activation within the kidney glomeruli.

To determine directly whether macrophages from PPARγ-KO and RXRα-KO mice were defective in apoptotic cell uptake we tested the ability of peritoneal macrophages to ingest fluorescently-labeled ATs in vitro. Compared with WT cells, PPARγ-KO and RXRα-KO peritoneal macrophages were markedly deficient in the phagocytosis of fluorescently labeled ATs (Fig. 6A, B), with a significantly lower average number of ATs scavenged per macrophage (phagocytosis index; Fig. 6C, 6D). This defect was also reflected in a reduced percentage of macrophages containing ingested ATs (Supplemental Fig. 7A, 7B). Phagocytosis of ATs by WT macrophages was significantly enhanced by pharmacologic activation of PPARγ or RXRα, increasing both the phagocytosis index (Fig. 6AD) and the percentage of WT cells containing ATs (Supplemental Fig. 7A, 7B). Treatment of PPARγ-KO or RXRα-KO macrophages with RSG or 9cRA had no effect on either parameter (Fig. 6C, 6D; Supplemental Fig. 7A, 7B). Real-time microscopy recordings indicated that PPARγ-KO and RXRα-KO macrophages were deficient in AT binding (Fig. 6E, 6F), and treatment with RSG or 9cRA strongly increased the AT binding ability of WT macrophages, but had no effect on macrophages lacking PPARγ or RXRα (Fig. 6G, 6H). Altered Fcgr1 and C1q expression was reflected by an impaired phagocytosis of Ig-opsonized (IgG and IgM, respectively) particles by PPARγ-KO and RXRα-KO macrophages (Supplemental Figs. 7D–G, 8). In contrast, macrophage deficiency in PPARγ or RXRα had no effect on the uptake of fluorescent beads (Supplemental Fig. 7C).

Our findings indicate that macrophage PPARγ and RXRα are required for the efficient disposal of apoptotic bodies, which prevents the release of intracellular components and the consequent loss of immunological self-tolerance, and might explain the development of an SLE-like nephritis in mice lacking macrophage PPARγ or RXRα. Indeed, altered PPARγ function has been linked to other autoimmune diseases, such as multiple sclerosis in humans and autoimmune encephalomyelitis in mouse (43). Our results show that the mechanism underlying the phagocytosis deficit of PPARγ-KO and RXRα-KO macrophages involves the disruption of a network of genes regulated by PPARγ/RXRα heterodimers. This gene network includes the cell surface receptors Cd36, Mertk, Axl, and Fcgr1, all of which are required for the proper binding and consequent internalization of apoptotic cells (9); Tgm2, which encodes a protein involved in apoptotic cell uptake (44); and the opsonins C1qa, C1qb and C1qc, which bind to the apoptotic cell surface to initiate phagocytosis (45). This coordinated regulation of phagocytosis-related genes indicates an important role for PPARγ/RXRα heterodimers in the clearance of apoptotic cells. However, the ability of the RXRα ligand LG268 to induce limited expression of these genes in PPARγ-KO macrophages suggests that they can be also regulated by other heterodimer partners of RXRα. Consistenly, two of these genes, Mertk and C1qb, have been shown to be targets of liver X receptor or PPARδ (39, 46). These findings indicate that RXRα can modulate the expression of phagocytosis-related genes by forming heterodimers with different partners.

There are considerable data supporting the idea that deficits in the expression of macrophage receptors and opsonins can lead to deficient apoptotic cell clearance and autoimmunity (11, 38, 39, 47). However, some mouse models of defective apoptotic cell removal do not have autoimmune disease (7, 48), suggesting that additional mechanisms contribute to the development of self immunity. The lack of the opsonin C1q in mice leads to altered apoptotic cell uptake and autoimmune disease with glomerular inflammation, which is similar to our present findings (13). Moreover, homozygous deficiency for C1q is associated with increased susceptibility to SLE in humans (13). There is also evidence that lack of MERTK and AXL results in widespread accumulation of apoptotic cells and autoantibody production in mice (49). Loss of MERTK alters macrophage cytokine production and impairs phagocytosis of apoptotic cells, leading to the development of a progressive SLE-like autoimmunity (50). Therefore, the lack of expression of these macrophage receptors and opsonins provides a possible molecular explanation for the deficient engulfment of apoptotic cells and the development of GN in PPARγ-KO and RXRα-KO mice.

Macrophages phagocytosing apoptotic cells accumulate cellular components, including cholesterol and fatty acids, that are potential activators of PPARγ/RXRα (51). Our results suggest that the activation of PPARγ/RXRα in this context has distinct effects on phagocytosis and the preservation of self-tolerance. First, PPARγ/RXRα activates the transcription of genes involved in phagocytosis, thereby mediating the rapid and early removal of dying cells and avoiding accumulation of their harmful intracellular components. Second, the PPARγ/RXRα mediated safe removal of apoptotic cells, which provokes a shift to a so-called regulatory or deactivated anti-inflammatory macrophage phenotype characterized by enhanced Tgfb1, Il10, Slam, and Mrc1 gene expression (8, 42) as well as reduction of proinflammatory genes such as Il1b, Il12b, Il6, Nos2 and Tnf. The inhibition of proinflammatory genes may be due to a direct transrepression activity of PPARγ by the ligand-dependent antagonism of the NF-κB (15). Apoptotic cells have been shown to induce PPARγ sumoylation, which is targeted to the nuclear receptor corepressor, blocking its clearance from NF-κB sites within proinflammatory promoters (52). In addition, the PPARγ/RXRα-mediated induction of Mertk and Axl may negatively regulate inflammatory mediators by controlling the expression of the suppressor of cytokine signaling proteins SOCS1 and SOCS3 (50).

Approaches aimed at resolving the core autoimmune disorder in SLE are at various stages of preclinical and clinical development, and they include immunosuppressive therapy and eventually hematopoietic stem cell transplantation (1). Ligands for PPARγ, such as RSG and pioglitazone (thiazolidinediones) can reduce renal inflammation and macrophage infiltration of kidney cortex in murine models of SLE (53), mainly through the modulation of mesangial cell functions (53, 54). Activation of PPARγ has also been proposed to ameliorate SLE nephritis through the upregulation of the synthesis of the anti-inflammatory factor adiponectin in adipocytes (55). Thiazolidinediones are already in clinical use as insulin sensitizers with antiatherogenic effects (56). This represents an advantage in regard to other nuclear receptors such as PPARδ and liver X receptor, whose ligands have not been introduced to clinical practice. Moreover, the overlapping effects of PPARγ and RXRα ligands on phagocytosis raise the possibility that dual administration of PPARγ and RXRα activators could allow the administration of lower doses of thiazolidinediones, reducing their side effects (56). Our work suggests a new therapeutic potential for PPARγ/RXRα ligands as enhancers of apoptotic cell removal by macrophages. Therefore, we anticipate that our findings will be a starting point for studies aimed at the selective modulation of macrophage PPARγ/RXRα potential targets for the treatment of SLE.

We thank Dr. Christian Hellriegel and Dr. Rubén Mota (Centro Nacional de Investigaciones Cardiovasculares [CNIC]), Roisin Brid Doohan (Histopathology Unit, CNIC), and Francisco Urbano Olmos and Covadonga Aguado Ballano (Transmission Electron Microscopy Laboratory, Universidad Autónoma, Madrid, Spain) for help and advice. We also thank Simon Bartlett for editorial assistance. The optical microscopy work described in this study was conducted at the Microscopy and Dynamic Imaging Unit, CNIC, Madrid, Spain.

Disclosures The authors have no financial conflicts of interest.