CD47 promotes autoimmune valvular carditis by impairing macrophage efferocytosis and enhancing cytokine production

Abstract

Systemic autoantibody-mediated diseases accelerate chronic cardiovascular disease in humans. In the K/B.g7 mouse model of spontaneous autoantibody-mediated inflammatory arthritis, valvular carditis arises in part due to Fc receptor-mediated activation of macrophages, leading to production of pathogenic TNF and IL-6. Here we explored whether impaired efferocytosis mediated by the interaction of CD47-expressing apoptotic cells with SIRPα on macrophages contributes to disease progression in this model. CD47-expressing apoptotic cells and SIRPα+ macrophages were abundant in inflamed/rheumatic cardiac valves from both mice and humans. In vivo anti-CD47 blockade both prevented and treated valvular carditis in K/B.g7 mice. Blocking CD47 enhanced macrophage efferocytosis and reduced macrophage production of TNF and IL-6. These studies highlight the CD47:SIRPα interaction as a key driver of chronic cardiac valve inflammation and suggest these molecules as potential therapeutic targets to reduce cardiovascular disease risk in autoantibody-driven inflammatory diseases.

Introduction

Macrophage engulfment and disposal of apoptotic cells, or efferocytosis, has been shown to play a critical role in the development of chronic inflammatory and fibrotic diseases. During chronic inflammation, increased apoptosis and impaired phagocytosis lead to accumulation of cellular debris. This overwhelms the efferocytotic capacity of macrophages, leading to persistent activation and exacerbated inflammation and fibrosis.

The anti-phagocytic, ‘don’t-eat-me’ molecule CD47 is upregulated on parenchymal cells during chronic inflammatory responses (1-5). CD47 inhibits phagocytic engulfment by interacting with signal regulatory protein alpha (SIRPα) expressed primarily on macrophages. In experimental settings, anti-CD47 and/or anti-SIRPα monoclonal antibody (mAb) blockade has shown therapeutic promise for chronic fibrotic and inflammatory diseases including atherosclerosis and cancer (1-10).

In the K/B.g7 TCR transgenic mouse model, chronic systemic inflammation arises due to linked T- and B-cell recognition of the ubiquitously-expressed autoantigen glucose-6-phophate isomerase (GPI), driving production of high-titers of anti-GPI autoantibodies (11-14). Anti-GPI autoantibody-containing immune complexes induce arthritis with a predictable timing of onset and trajectory. In addition to arthritis, K/B.g7 mice develop fibro-inflammatory valvular carditis with many similarities to human inflammatory valve disease (15, 16). This model is therefore useful to dissect the mechanisms by which chronic autoantibody-associated rheumatic diseases such as rheumatic heart disease, rheumatoid arthritis and systemic lupus erythematosus promote cardiovascular disease progression (17).

Our group has previously shown that discrete populations of macrophages expressing macrophage galactose-type C-type lectin 2 (MGL2/CD301b) and CX3CR1 are key mediators of mitral valve inflammation in the K/B.g7 model (16). Circulating immune complexes bind to activating Fc gamma receptors on these macrophages and initiate signaling through the cytoplasmic tyrosine kinase Syk (spleen tyrosine kinase), resulting in pro-inflammatory TNF and IL-6 production (16, 18).

Here we investigated whether the CD47:SIRPα pathway was at play in the K/B.g7 model of valvular carditis. We find that CD47-expressing apoptotic cells accumulate in chronically inflamed mitral valves (MV). We report that anti-CD47 monoclonal antibody treatment both prevents and treats valve inflammation. Importantly, blockade of CD47 improves efferocytosis and reduces macrophage production of TNF and IL-6. Finally, we demonstrate correlative findings in cardiac valve samples from patients with chronic rheumatic heart disease (RHD). Collectively, these findings support a working model in which pathogenic production of TNF and IL-6 by macrophages is driven both by immune complex engagement of activating FcRs and by CD47:SIRPα-mediated inhibition of efferocytosis.

Materials and Methods

Animals

Animal studies were approved by the University of Minnesota Institutional Animal Care and Use Committee (IACUC protocols 2102-38871A, 1805-35885A, 1506-32700A and 1207A17481). Standard specific pathogen free (SPF) housing was used. KRN T cell receptor (TCR) transgenic (‘KRN’) and C57BL/6:I-Ag7 MHC class II congenic (‘B.g7’) were gifts from D. Mathis, C. Benoist, and the Institut de Génétique et Biologie Moléculaire et Cellulaire (IGBMC), Illkirch-Graffenstaden, France. C57Bl/6 mice (stock # 000664) were purchased from The Jackson Laboratory. Male and female mice were used in roughly equal proportions throughout the study.

Monoclonal antibody (mAb) blockade studies

Monoclonal antibody anti-CD47 (clone MIAP410) for in vivo use was purchased from BioXcell. Isotype control mouse IgG was purchased from Jackson Immunoresearch, Inc. Antibodies were diluted to 1 mg/ml in phosphate-buffered saline (PBS) and 0.2 ml (200 μg) was injected intraperitoneally twice weekly.

Arthritis progression

Clinical arthritis scoring was conducted using a 12-point system described previously (19) in which each paw is assigned a score of 0 to 3. In this system, 0 denotes absent joint inflammation and 3 denotes maximal swelling. Additionally, changes in mouse ankle thicknesses over the course of each experiment were determined using a method described previously (14), using a dial gauge (Käfer, model J15).

Determination of anti-glucose-6-phosphate isomerase (anti-GPI/α-GPI) antibody titers

Anti-GPI antibody serum titers were quantified using enzyme-linked immunosorbent assay (ELISA) and quantified colorimetrically using a microplate reader (BioRad iMark) as described previously (13, 20). Samples were subsequently normalized as follows for comparisons between groups and across experiments: each individual absorbance measurement was divided by the average absorbance for each control group of samples within each experiment.

Mouse tissue histology

Following CO₂ euthanasia, mice were perfused with 10 ml chilled PBS; hearts were dissected and placed in Optimal Cutting Temperature (OCT) medium (Thermo Fisher Scientific). Hearts were immediately flash-frozen over a dry ice and 2-methyl-butane slurry and stored at −80°C. Frozen cardiac tissue was sectioned using a cryostat (Leica CM3050 S). Coronal sections were cut at 7 μm thickness and transferred to Superfrost™ Plus glass slides (Thermo Fisher Scientific). Hematoxylin and eosin (H&E) staining was conducted according to standard methods. Masson’s trichrome staining was conducted according to the manufacturer’s recommendations (Abcam). Histologic sections were used to measure mitral valve thicknesses as described previously (15). CellSens image acquisition software (Olympus) and an Olympus BX51 microscope were used for brightfield image acquisition.

Immunofluorescent staining of mouse tissue

Sections from flash-frozen mouse hearts were prepared for immunofluorescent (IF) staining as described previously with minor modifications (16). Briefly, sections were removed from storage and allowed to warm to room temperature for 15 minutes and then fixed in cold 4% paraformaldehyde diluted in PBS for 10 minutes. After washing, each section was permeabilized using 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 10 minutes and then washed thrice in PBS for five minutes. Non-specific binding was blocked for 1 hour during which each section was incubated in PBS containing 5% normal donkey serum (NDS, Jackson Immunoresearch Laboratories, Inc), 2% bovine serum albumin (BSA), and 0.1% Tween-20 (Sigma Aldrich, catalog # P1379). Endogenous Fc receptors were blocked using 5 μg/ml anti-CD16/32 (rat monoclonal [2.4G2], Tonbo Biosciences) added to the blocking buffer. The following fluorochrome-conjugated antibodies were employed for direct IF: Alexa Fluor-488 (AF-488)-conjugated anti-active-caspase-3 (anti-casp3, rabbit polyclonal [Cell Signaling Technology); AF-647-conjugated anti-CD47 (clone miap301 [Biolegend]); AF-594-conjugated anti-SIRPα (clone, P84 [Biolegend]). Control samples were processed in parallel using species-matched isotype antibodies (rabbit polyclonal IgG AF-488 conjugated [Cell Signaling Technology]; AF-647 conjugated rat IgG2aκ, clone RTK2758 [Biolegend]; AF-594 conjugated rat IgG1κ, clone RTK2071 [Biolegend].

After antibody staining, each section was incubated for 10 minutes in PBS containing 1 μg/ml Hoechst 33342 (Thermo Fisher Scientific) to counterstain nuclei. Each sample was washed thrice for 5 minutes in PBS. Coverslips were mounted using ProLong Gold Antifade fluorescence mounting medium (Thermo Fisher Scientific). Images were acquired within 24 hours following mounting using a Leica DM6000B epi-fluorescence microscope and LAS X software (Leica) for image acquisition. Exposures were matched to negative staining in species-matched isotype control samples. Contrast and brightness were enhanced 20% in some panels, a change applied uniformly across samples within the same experiment.

Mouse mitral valve isolation and digestion

Murine mitral valves were isolated as described previously (16). Briefly, after euthanasia, hearts were perfused with 5 ml of chilled PBS. Excised hearts were placed in 0.5 ml ice cold RPMI medium containing GlutaMAX supplement (Thermo Fisher Scientific), and an additional 3% fetal bovine serum (FBS, Thermo Fisher Scientific), and 10 mM HEPES, Thermo Fisher Scientific). Valve tissue was excised with the aid of a dissecting microscope and placed into 1 ml of the same medium described above plus 500 U/ml collagenase-2 (Worthington Biochemical) and 20 U/ml DNase I (Worthington Biochemical), pre-warmed to 37°C. Valve tissue samples were incubated at 37°C for 1-hour with intermittent trituration using a P-1000 pipette. Samples were centrifuged for 5 minutes at 300xg at 4°C. Pellets were washed once with PBS containing 2% w/v bovine serum albumin (PBS/2% BSA). Erythrocytes remaining in the pellet were lysed using 0.1 ml ammonium chloride-potassium (ACK) lysis buffer containing 150 mM NH₄Cl, 10 mM KHCO₃, 0.10 mM EDTA, followed by incubation at room temperature for 3 minutes and quenched using 1.5 mL cold PBS/2% BSA. The samples were centrifuged at 300xg RCF for 5 minutes at 4°C. The remaining pellets were re-suspended in 1 ml PBS/2% BSA; undigested cell and tissue debris was removed by filtering through 70-micron cell strainer caps (Thermo Fisher Scientific). After centrifuging the filtered cell suspensions at 300xg RCF for 5 minutes at 4°C, the remaining cell pellets were re-suspended in 0.2 mL cold PBS/2% BSA, transferred to 96-well plates for staining for flow cytometry.

Preparation and flow cytometry of mouse cardiac tissues

The following antibodies and clones were used: CD3ε (clones 145-2C11 and 17A2), CD45R/B220 (clone RA3-6B2), CD45.2 (clone 104), Ly6G (clone 1A8), Ly6C (clone AL-21), were purchased from BD Biosciences; GR-1 (clone RB6-8C5), CD11b (clone M1/70), F4/80 (clone BM8) purchased from eBioscience, CD31 (clone 390), CX3CR1 (clone SA011F11), CD301b/MGL2 (clone URA-1), CD64 (clone X54-5/7.1), TNF (clone MP6-XT22), IL-6 (clone MP5-20F3), CD47 (clone miap301 and CC2C6), SIRPα (clone P84) were purchased from Biolegend; Feeder Cells (clone mEF-SK4) purchased from Miltenyi Biotec; active/cleaved caspase-3 (rabbit polyclonal) purchased from Cell signaling Technologies. Live cells were identified based on viability dye exclusion (Ghost Dye Violet 510, Tonbo Biosciences) used according to the manufacturer’s recommendations. Fc receptors were blocked using anti-CD16/32 (α-FcγRII/III [2.4G2], Tonbo Biosciences) at 5 μg/ml. For intracellular antigens (e.g. TNF, IL-6, active/cleaved-caspase3; specific methods described below), a commercially available kit for fixation and permeabilization was employed according to the manufacturer’s recommendation (BD). Fluorochrome compensation matrices were generated using single stain controls and flow cytometry compensation beads (UltraComp eBeads Compensation Beads, Thermo Fisher Scientific). Negative and positive gates were assigned using fluorescence-minus-one (FMO) controls. LSR Fortessa and LSR II cytometers (BD Biosciences) equipped with FACSDiva software for data acquisition were used. Data were analyzed using FlowJo version 10.

Intracellular cytokine staining

For quantification of intracellular cytokine production in mouse cells using flow cytometry, each mouse was injected with 0.25 mg brefeldin-A (Biolegend) intraperitoneally 3 hours prior to euthanasia. Hearts were isolated as described above. An additional 1 μg/ml of brefeldin-A was added to all buffers used for tissue processing to enhance intracellular cytokine detection. Tissue processing and cell surface staining of each sample were conducted as described above. Following cell surface staining, leukocyte samples were fixed and permeabilized using buffer containing 3.7% paraformaldehyde and saponin (BD Biosciences Fixation/Permeabilization buffer). Anti-TNF, IL-6, and active/cleaved caspase-3 (listed above) were diluted in staining buffer containing fetal bovine serum (FBS) and 0.083% sodium azide Stain Buffer, [BD Biosciences]) and incubated for 45 minutes at room temperature. Each sample was washed twice and filtered through a 70-micron prior to flow cytometry analysis.

CaspGLOW fluorescence staining

Following cell transfer to 96-well plate, cells were resuspended in 300 μL cold RPMI medium containing GlutaMAX supplement (Thermo Fisher Scientific), 10 mM HEPES, and an additional 10% heat-inactivated FBS (Thermo Fisher Scientific). Following resuspension, 1 μL of FITC-VAD-FMK (CaspGLOW, Thermo Fisher Scientific) was added to each sample. Non-fluorescent Z-VAD-FMK was used as a negative control. Samples were then incubated for 45 minutes in a 37°C incubator with 5% CO₂. Samples were centrifuged at 300xg RCF for 5 minutes at 4°C and washed twice with CaspGLOW wash buffer (Thermo Fischer Scientific). Cells were resuspended in PBS and washed before proceeding with viability dye staining, Fc-blockade, and extracellular staining as described above.

Image analysis

The Fiji distribution of ImageJ (NIH) was used for image analysis. The image analyses are based on methods described previously (16). Briefly, species-matched isotype control samples were analyzed in parallel and used to assign fluorescence exposures. A 40x objective lens was used to acquire between 5-10 focal planes within each field of view and tile-scan mode was employed to acquire images from the entirety of each valve. Greyscale maximum-intensity z-projections were generated for each channel of interest. Area-normalized staining intensity was determined using these maximum-intensity projections after converting to 8-bit format.

Quantification of fluorescent signal as a percentage of valve area was accomplished by outlining the valve leaflets within each field of view to create a region of interest (ROI). All images in the channel of interest were then converted to a binary image using the threshold feature in FIJI. Threshold level was set based on the signal in the isotype control to ensure positive fluorescent signal was representative of the antigen of interest. The ROI valve tracings were overlayed on each respective binary image, and the positive signal in the ROI area was measured using the analyze feature in FIJI. The signal area as a percentage of total valve ROI area is reported. 2 to 3 heart sections per biological replicate were quantified and averaged to represent one replicate before statistical analysis.

Serum Cytokine Analysis

Serum cytokine analysis was performed using the mouse specific cytokine/chemokine panel on the Luminex platform as a multi-plex. The magnetic bead set (cat. #MCYTMAG-70K-10C) was purchased from EMD Millipore, St. Charles MO. Samples were assayed according to manufacturer’s instructions. Fluorescent color-coded beads coated with a specific capture antibody were added to each sample. After incubation, and washing, biotinylated detection antibody was added followed by phycoerythrin-conjugated streptavidin. Assay was read on a Luminex instrument (MAGPIX). Samples were run in duplicate and values were interpolated from 5-parameter fitted standard curves.

Human valve tissue procurement

Human mitral valve tissue samples from RHD patients undergoing surgical correction were acquired during missions in countries with high disease incidence. Harvesting followed University of Minnesota patient consent procedures in adherence with IRB 1307M39481. Due to the de-identified nature of the samples, histological analyses were deemed exempt from IRB oversight. Control human mitral valve tissue was acquired at the University of Minnesota Medical Center in accordance with IRB STUDY00003871.

Human valve tissue immunofluorescent (IF) staining

Tissue preservation following surgical resection and preparation for IF staining assessment was performed as described previously (16). Formalin-fixed, paraffin-embedded (FFPE) sections were rehydrated using a decreasing ethanol/water gradient. Antigen retrieval was accomplished by incubating each section in 95°C citrate buffer, pH 6.0, for 30 minutes followed by a 20-minute cool-down period. Permeabilization was accomplished by incubating in PBS containing 0.1% Triton X-100 for 20 minutes. Fc receptor blocker (Innovex) was applied for 30 minutes at room temperature according to the manufacturer’s recommendation. Following Fc blocking, non-specific binding was blocked by incubating each section for one hour in PBS containing 5% NDS, 2% BSA, 0.1% Tween-20. Primary antibody incubation was performed overnight at 4°C in the same buffer used for blocking (PBS containing 5% NDS, 2% BSA, and 0.1% Tween-20). The following primary antibodies were employed: anti-CD47 (mouse monoclonal [B6H12.2], Thermo Fisher Scientific), active-caspase-3 (rabbit monoclonal [5A1E], Cell Signaling Technology). Isotype control antibodies were used in parallel at the same concentrations: rabbit IgG (Vector Laboratories), mouse IgG and goat IgG (Jackson Immunoresearch). The following species-matched secondary antibodies were purchased from Jackson Immunoresearch and used for detection: Alexa Fluor 594 (AF-594)-conjugated donkey-anti-mouse, Alexa Fluor 647 (AF-647)-conjugated donkey-anti-rabbit. For improved fluorescence resolution from each of these channels, a modified filter set was employed (excitation maximum/emission maximum): 585/625 nm (Alexa Fluor 594, AF-594), 620/700 nm (Alexa Fluor 647, AF-647).

Bone marrow-derived macrophages

To generate bone marrow-derived macrophages, bone marrow was flushed from the femurs and tibias of a mouse’s hind limbs. Cells were treated with AKC lysis buffer to lyse red blood cells and then filtered through a 70 μm filter. Bone marrow cells were plated on 15-cm, untreated culture dishes (Corning, 430597) at 10x10⁶ cells per dish in DMEM supplemented with 10% CMG14-12 conditioned media (CM), 10% fetal bovine serum (FBS, Omega Scientific, FB-11), 1 mM sodium pyruvate (Gibco, 11360070), 2 mM L-glutamine (Gibco, 25030081), and 100 U/mL Penicillin-Streptomycin (Gibco, 15140148). Cells were left to differentiate and proliferate for 6 or 7 days. Cells were then passaged for treatments using cell dissociation media (Gibco, 13151-014) and maintained in the aforementioned media without CMG14-12 CM supplementation.

Jurkat cells

Jurkat cells were cultured in RPMI media (Gibco, 11875-093) supplemented with 5% FBS (Omega Scientific, FB-11), 2 mM L-glutamine (Gibco, 25030081), and 100 U/mL Penicillin-Streptomycin (Gibco, 15140148) at a maximum density of 1.2 x 10⁶ cells per mL. Apoptosis was induced by exposing the cells to UV light (254 nm) for five minutes (unless otherwise specified in figures) in culture media.

BMDM treatments

BMDMs were plated in 6 well plates (Corning, 3736) at 1x10⁶ cells per well and allowed to adhere overnight. For fluorescence imaging experiments, a sterilized 22 mm x 22 mm glass coverslip was placed in each well prior to addition of cells.

The next day, Jurkat cells were stained with PKH67 (Sigma-Aldrich, MINI67-1KT) following manufacturer’s instructions. Apoptosis was induced as described above. Jurkat cells were then treated with anti-CD47 monoclonal antibody (ThermoFisher, MA5-11895, clone B6H12.2) or isotype control (ThermoFisher, mouse IgG1κ, 14-4714-82) at varying concentrations (see figures) and incubated at 37°C, 5% CO₂ for one hour. BMDMs were pre-treated with Fc blocking antibodies (Biolegend, CD16, clone S17014E; CD16.2, 9E9; CD64, X54-5/7.1) or Fc isotype control (Biolegend, IgG2aκ, RTK2758; AHIgG, HTK888; mouse IgG1κ, MOPC-21) for one hour at 37°C, 5% CO₂. Apoptotic Jurkat cells (1x10⁶/mL) were then added to BMDMs cultures (0.5x10⁶/mL) and incubated at 37°C, 5% CO₂ for 6 or 24 hours (see figures).

BMDM fluorescence imaging

Following treatments, Jurkat/ BMDM cultures were gently washed with PBS and then fixed with 4% paraformaldehyde in PBS for 5 minutes at room temperature. Cultures were again washed with PBS, permeabilized with 0.1% triton-X, washed with PBS, and then stained with 2 μg/mL DAPI for 5 minutes at room temperature. Following one more PBS wash, coverslips were gently removed from each well and mounted cell-side down on microscope slides (FisherScientific, 12-550-15) using mounting media (Vector Labs, Vectashield). 10x images were captured on an Olympus BX51 microscope. Images were processed and quantified using ImageJ software.

BMDM and Jurkat flow cytometry

Following treatments, cells were washed with PBS. BMDMs were removed from culture wells using cell dissociation media (Gibco, 13151-014) and washed with 2% BSA in PBS. Cells were then resuspended in Fc Block (5 μg/mL) on ice for 5 minutes. BMDMs were stained with fluorescently conjugated antibodies against macrophage-specific surface proteins including SIRPα and Jurkat cells were stained with primary anti-CD47 monoclonal antibody (ThermoFisher, MA5-11895, clone B6H12.2) or isotype control mouse IgG1,κ (ThermoFisher, 14-4714-82) on ice for 20 minutes in the dark. Jurkats were then stained with goat anti-mouse IgG-Texas Red (JacksonImmuno, 115-076-071) for 15 minutes on ice. All cells were washed with PBS and resuspended in a Ghost viability dye (Tonbo Biosciences) for 20 minutes in the dark. To determine the extent of apoptosis induced in Jurkats, cells were stained with Annexin V (ThermoFisher, A13201) and propidium iodide following manufacturer’s instructions. Briefly, cells were resuspended in annexin-binding buffer and AF488-conjugated annexin V was added to each sample and incubated for 15 min at room temperature. Stain was then diluted in more annexin-binding buffer and propidium iodide was added to samples. Cells were run on a cytometer soon after staining.

Statistics

Graphpad (Prism) was used for all statistical analyses. Non-parametric, two-tailed Mann-Whitney tests were used to compare cardiac valve thickness measurements, normalized anti-GPI serum titers, immunofluorescent signal area of MV, serum cytokine levels, and flow cytometry data on MVs from untreated mice. Flow cytometry data on α-CD47 treated mice were compared using parametric two-tailed student’s t-tests. Statistical analysis of arthritis scores and changes in ankle thicknesses were assessed using two-way repeated measures ANOVA with post-hoc Sidak tests for multiple comparisons. Statistical analysis for in vitro Jurkat uptake by BMDMs used one-way ANOVA followed by Kruskal-Wallis test of multiple comparisons. Sample sizes (n) refer to biological replicates, listed in the figures and/or legends. Statistical significance was defined as p<0.05, and indicated as *p<0.05, **p<0.01, ***p<0.005, ****p<0.0001.

Results

Macrophages are critical mediators of valvular carditis and fibrosis in the autoantibody-dependent K/B.g7 mouse model. We previously showed that mitral valve inflammation in K/B.g7 mice depends on activating Fc receptors, and that downstream FcR signaling in macrophages through the cytoplasmic tyrosine kinase Syk drives pathogenic TNF and IL-6 production (16, 18). We hypothesized that additional macrophage activation pathways might also contribute. Based on the knowledge that macrophage-mediated clearance of apoptotic cells (efferocytosis) is essential for normal cardiac valve development (21) and that impaired efferocytosis can promote other inflammatory cardiovascular diseases (e.g. atherosclerosis) (7), we explored the role of apoptotic cell clearance in K/B.g7 valvular carditis, including determining whether these pathways impinge on macrophage cytokine production.

We focused specifically on the CD47:SIRPα pathway. CD47 is a “don’t-eat-me” signal expressed by apoptotic cells that protects them from efferocytosis by interacting with the inhibitory receptor SIRPα on macrophages. We asked whether apoptotic cells were a feature of K/B.g7 mitral valve disease. Using flow cytometry, we investigated apoptotic cell burden and CD47 expression relative to non-inflamed control mitral valves. We used a reagent that irreversibly binds to active caspase enzymes of living cells, allowing identification of cells in any stage of caspase-mediated apoptosis. The gating scheme is shown in Supplemental Figure 1A. Relative to control valves, inflamed K/B.g7 mitral valves had significantly more apoptotic cells, which also comprised a greater proportion of total cells (Figures 1A and 1B). Additionally, CD47 expression was higher on apoptotic cells from inflamed K/B.g7 valves than on non-inflamed control valves, and higher on apoptotic cells than on non-apoptotic cells (Figure 1C). We further characterized the types of cells undergoing apoptosis and found that they were a heterogeneous population (Figure 1D, top row). Specifically, apoptotic cells in the K/B.g7 mitral valves (MVs) were predominantly immune cells, whereas fibroblasts comprised the largest proportion of apoptotic cells in non-inflamed valves (Figure 1D). It is possible that some of these immune cells are macrophages which have engulfed other types of apoptotic cells. To complement these flow cytometric studies, we analyzed valves by immunostaining. We found that CD47 and SIRPα were both abundantly expressed in inflamed K/B.g7 mitral valves (Figure 1E). Additionally, inflamed mitral valves contained foci of active-caspase-3⁺ apoptotic cells distributed throughout the valve interstitium, along with abundant CD47 and SIRPα expression (Figure 1E). We observed similar findings in valves from patients with RHD, including accumulation of apoptotic/active-caspase-3⁺ cells and increased expression of CD47 across multiple patient samples (Supplemental Figure 2) relative to non-inflamed human mitral valve tissue (data not shown). Thus, CD47 and apoptotic/active-caspase-3⁺ cells are easily detected in inflamed mitral valves from K/B.g7 mice and human RHD patients. Based on these observations, we hypothesized that CD47 expression impairs efferocytosis and contributes to inflammatory mitral valve disease.

Figure 1. Apoptosis and CD47 expression are enhanced in inflamed K/B.g7 mitral valves.

(A) Flow cytometric quantification of total apoptotic/caspase-expressing mitral valve (MV) cells (gated on single, live cells). Caspase⁺ cells identified by CaspGLOW (Thermo) staining. (B) Frequency of caspase-expressing cells as a percentage of total single, live MV cells. (C) Mean fluorescence intensity of CD47 on single, live, caspase-negative versus caspase-positive cells from the indicated mice. (D) Pie charts display identified cell types (right) as a proportion of caspase⁺ positive (top row) and caspase⁻ (bottom row) MV cells from B6 (left column) and K/B.g7 (right column) mice. Hematopoietic (CD45.2⁺) subsets further gated onT cells (CD3ε⁺B220⁻), B cells (B220⁺CD3ε⁻), neutrophils (B220⁻CD3ε⁻GR-1⁺), macrophage populations (CD3ε⁻B220⁻GR⁻ CD11b⁺CD64⁺), and other (CD45.2⁺CD3ε⁻B220⁻GR⁻CD64⁻CD11b⁻); non-hematopoietic (CD45.2⁻) subsets are gated as fibroblasts (mEF-SK4⁺CD31⁻), endothelial cells (ECs; CD31⁺mEF-SK4⁻), co-expressing cells (fibro/ECs; mEF-SK4⁺CD31⁺), and other cells (CD45.2⁻mEF-SK4⁻CD3⁻). Graphs are representative of at least 96% of total single, live MV cells in each group. (A-D) Gating scheme depicted in Supplemental Fig 1B. MV samples displayed from 8-week-old non-inflamed B6 mice (n=5) and 10-week-old inflamed K/B.g7 mice (n=6). Data analyzed using an unpaired, nonparametric, two-tailed Mann-Whitney test and represented as mean ± SD. (E) Immunofluorescent (IF) staining of 9-week-old B6 (left) and K/B.g7 (right) MV for SIRPα [red], active-caspase-3 (Casp3) [green], CD47 [blue], and nuclei [white]. Individual channels (top) displayed above merged image. Increased magnification (bottom, scale bar = 100 μm) of boxed area is displayed under lower magnification (scale bar = 50 μm). LA indicates left atrium; LV, left ventricle.

We used anti-CD47 blocking antibodies to test this hypothesis. First, a preventative regimen was initiated at the onset of clinical disease (4 weeks of age). Second, a therapeutic regimen was initiated at 6 weeks of age, a timepoint when mitral valve inflammation and fibrosis are well-established (Figure 2A) (15, 18, 22). We found that preventative CD47 blockade significantly reduced mitral valve inflammation and fibro-inflammatory thickening, both qualitatively and quantitatively (Figure 2B). While preventative CD47 blockade reduced the severity of arthritis early in its course, by the conclusion of the treatment no difference in overall arthritis severity was observed and no differences in autoantibody titers were observed (Figure 2D). Similar effects were seen when anti-CD47 blockade was delivered therapeutically, after the onset of valve inflammation (Figure 2C and 2E). These findings demonstrate that anti-CD47 therapy both prevents and treats mitral valve inflammation and fibrosis in K/B.g7 mice, essentially normalizing mitral valve thickness. In addition, CD47 blockade significantly reduced expression of the endothelial activation marker VCAM-1 (Figure 2F and G). In valves from mice treated with isotype control antibody, the VCAM-1 signal was diffuse, a sign of disorganized remodeling. In contrast, in the setting of CD47 blockade VCAM-1 expression was confined to the endothelial layer. This reduction in VCAM-1 expression could be due to direct effects of anti-CD47 blockade or, more likely, to secondary effects of decreased cytokine production, as we have previously shown that blocking IL-6 or TNF similarly reduces VCAM-1 expression (16). Overall, these experiments demonstrate that CD47 blockade reduces mitral valve inflammation.

Figure 2. CD47 blockade prevents and treats MV inflammation in K/B.g7 mice.

(A) Schematic timeline for preventative (orange) and therapeutic (blue) anti-CD47 (α-CD47) monoclonal antibody (mAb) blockade or mouse IgG (mIgG) isotype control in K/B.g7 mice. (B) and (C) Representative Masson’s trichrome stained MV sections from mIgG (top panel) or α-CD47 (bottom panel) treated K/B.g7 mice at 8 weeks (preventative, B) and 10 weeks (therapeutic, C). Thickness quantification of MV sections from K/B.g7 mice treated with preventative or therapeutic α-CD47 blockade or mIgG. (D) and (E) average change in ankle thickness (left) in α-CD47 or mIgG blockade during indicated treatment. Colored regions indicate the treatment period. Normalized anti-GPI IgG serum titers from indicated treatment groups (right). (F) IF staining of VCAM-1 [yellow] and nuclear stain [blue] in MV [white outline] sections from K/B.g7 mice treated with preventative mIgG (left) and α-CD47 (right) treated. (G) Quantification of VCAM-1 signal area as a percentage of total MV area following preventative dosing (α-CD47 n=5, mIgG n=5). Arthritis was assessed with a 2-way repeated measures ANOVA with Sidak’s multiple comparisons. All other data analyzed using a two-tailed Mann-Whitney test. Data represented as mean ± SD. *p<0.05, **p<0.01, ***p<0.005 at specified timepoint; ns, not significant at any timepoint. Scale bars = 50 μm. AU indicates arbitrary units; LA, left atrium; LV, left ventricle.

CD47 expression promotes accumulation of apoptotic cells in K/B.g7 mitral valves

Because CD47 provides a “don’t-eat-me” signal to SIRPα-expressing macrophages, we hypothesized that antibody blockade of CD47 would promote macrophage uptake of apoptotic cells in the mitral valves. Indeed, mitral valves from K/B.g7 mice treated preventatively with anti-CD47 blocking antibodies demonstrated qualitatively diminished active-caspase-3 staining relative to mice treated with isotype control antibody (Figure 3A). Additionally, in the setting of CD47 blockade, active-caspase-3 positive cells appeared to be contained within SIRPα positive cells (Figure 3B, arrowheads). Quantification of active-caspase-3⁺ signal across multiple images and samples revealed that CD47 blockade significantly reduced the overall apoptotic cell burden in the inflamed mitral valves, (Figure 3C).

Figure 3. CD47 blockade diminishes active-caspase-3+ cell accumulation.

(A) IF staining for SIRPα [red], active-caspase-3 (Casp3) [green], CD47 [white?], and nuclei [blue?] in MVs from preventative α-CD47 blockade (top panels: mIgG; bottom: α-CD47). (B) Increased magnification of mIgG and α-CD47 treated MV with active-caspase-3 and SIRPα channels overlayed. SIRPα⁺ cells co-expressing Casp3 [yellow] are indicated by white arrowheads. (C) Quantification of Casp3 signal area as a percentage of total MV area in K/B.g7 mice that received preventative treatment (α-CD47 n=5, mIgG n=4). Representative images displayed in (A). Relative Casp3 staining were analyzed using a two-tailed Mann-Whitney test. Scale bars = 50 μm. LA indicates left atrium; LV, left ventricle. Data are represented as mean ± SD.

The findings above suggest that the reduction in active-caspase-3⁺ cells during CD47 blockade is due to increased macrophage efferocytosis. However, we also considered the possibility that the reduction in active-caspase-3⁺ cells could be due to reduced immune cell recruitment and reduced apoptotic cell generation in the setting of prolonged CD47 blockade. To exclude this possibility, we used an acute CD47 blocking regimen initiated at the point of maximal mitral valve inflammation (8 weeks of age) and continued for only two weeks (Figure 4A). In these experiments, we reasoned that mitral valves in the period of maximal disease activity would already have accumulated apoptotic cells, so that we could then observe the effects of short-term CD47 blockade. We used flow cytometry to analyze the populations infiltrating the inflamed mitral valves. This regimen had no effect on arthritis progression or autoantibody production (Figure 4B). We verified that essentially all valve macrophages expressed SIRPα (Figure 4C). Although there was a trend toward fewer active-caspase-3⁺ cells in mice treated acutely with CD47 blockade (Figure 4D), this did not reach significance; thus, the acute anti-CD47 blocking protocol worked as intended. This acute anti-CD47 blocking regimen did significantly decrease the number of total leukocytes and CX3CR1⁺ macrophages in the inflamed mitral valves (Figure 4E and 4F); this could be due either to a direct effect of anti-CD47 blockade or an indirect effect of reduced inflammatory cytokine production, as shown below.

Figure 4. Acute CD47 blockade promotes apoptotic cell clearance via efferocytosis in K/B.g7 mitral valves.

(A) Schematic timeline for acute anti-CD47 (α-CD47) or control mouse IgG (mIgG) blockade in K/B.g7 mice. (B) Average change in K/B.g7 ankle thickness (left) preceding and during acute treatment period. Normalized anti-GPI IgG serum titers from acute treatment groups (right). (C) Flow cytometric quantification of SIRPα expression in mitral valve (MV) mononuclear phagocytes (MNPs) compared to MNPs in SIRPα FMO control sample (gated on live, single, Ly6G⁻B220⁻CD3ε⁻CD45.2⁺CD11b^hi/lowCX3CR1^hi/low). (D) Representative flow cytometric quantification (MFI) of intracellular active-caspase-3 (Casp3⁺) staining in inflamed MVs taken from K/B.g7 mice treated with acute mIgG or α-CD47; graph at right quantifies total Casp3+ cells (gated on single, live cells). (E) Flow cytometric quantification of total MV leukocytes (single, live, CD45.2⁺ cells) following acute treatment. (F) Flow cytometric quantification of MV CX3CR1⁺ MNPs following acute treatment (gated on single, live, Ly6G⁻B220⁻CD3ε⁻CD45.2⁺). (G) Flow cytometric separation of MV MNPs into CX3CR1^hi and CX3CR1^low populations. (H) Quantification of Casp3⁺ cells within live CX3CR1^hi MNP populations (left) and live CX3CR1^low MNP populations (right) following acute treatment. Gating scheme depicted in Supplemental Fig 1B. Flow cytometric data were analyzed using a two-tailed student’s t-test; arthritis metrics were assessed with a 2-way repeated measures ANOVA with Sidak’s multiple comparisons; anti-GPI titers normalized to mIgG treated mice and analyzed with a two-tailed Mann-Whitney test. NS indicates not significant. AU indicates arbitrary units. Red bars indicate mean ± SD.

We next asked whether interfering with the CD47:SIRPα interaction increased phagocytic cell uptake of apoptotic cells, as predicted. We designed a flow cytometric experiment to detect apoptotic, active-caspase-3-expressing cells within live CD11b⁺ macrophage populations. By staining for intracellular active-caspase-3 and gating out dead cells, we aimed to capture a snapshot of a small number of macrophages in the process of efferocytosis (gating scheme is shown in Supplemental Figure1B). Despite the reduction in total valve macrophage numbers, acute CD47 blockade increased the fraction of live CD11b⁺CX3CR1^high and CD11b⁺CX3CR1^low macrophages containing active-caspase-3⁺ apoptotic cells (Figure 4G and H). Although acute CD47 blockade reduced the number of valvular immune cells overall, the remaining CD11b⁺ CX3CR1^high/low macrophages appear more efficient at internalizing apoptotic cells. Collectively, these data suggest that CD47 blockade reduces mitral valve inflammation in K/B.g7 mice while enhancing macrophage-mediated efferocytosis of apoptotic cells.

We considered the possibility that the anti-CD47 antibody therapy was promoting efferocytosis via macrophage activating Fc receptor mediated uptake of CD47-expressing cells coated with anti-CD47 antibodies, in addition to or instead of by blocking the CD47:SIRPα interaction. We developed an in vitro assay to address this possibility (Supplemental Figure 3). We incubated irradiated, CD47-expressing human Jurkat T cells with mouse bone marrow-derived macrophages, using different combinations of anti-CD47 with antibodies that block the murine activating FcγRs (FcγR1, 3, and 4) and verified the expression of SIRPα on BMDMs (Supplemental Figure 3C). Importantly, it is known that mouse SIRPα can interact with human CD47 (23). At baseline, the BMDMs took up few of the irradiated cells. As anticipated, addition of anti-CD47 antibody increased uptake of the irradiated cells. The addition of FcR blocking antibodies reduced this effect of the anti-CD47 antibody, though not to baseline (Supplemental Figure 3D). Thus, the beneficial effect of the anti-CD47 antibody in vivo could be due to both of these mechanisms: blocking the CD47:SIRPα interaction and promoting FcR mediated uptake of CD47-expressing apoptotic cells.

Acute CD47 blockade attenuates IL-6 and TNF production by mitral valve-infiltrating macrophages

We have previously shown that TNF and IL-6 are key drivers of mitral valve disease in K/B.g7 mice (16). Based on a prior report that macrophage efferocytosis of apoptotic cells inhibits inflammatory cytokine production in vitro (24), we reasoned that the efferocytosis-promoting activity of CD47 blockade would similarly inhibit macrophage cytokine production in our in vivo model, providing a mechanistic link between enhanced efferocytosis and the reduced severity of valvular carditis. Indeed, acute CD47 blockade reduced mitral valve SIRPα-expressing macrophage production of TNF and IL-6, calculated both as the absolute number of macrophages producing these cytokines as well as per-cell cytokine expression (mean fluorescence intensity) (Figure 5).

Figure 5. Acute CD47 blockade attenuates mononuclear phagocyte TNF and IL-6 production in inflamed K/B.g7 mitral valves.

(A) Detection of intracellular TNF and IL-6 production by MV MNPs following acute blockade of α-CD47 (red) or mIgG (black) in K/B.g7 mice (left). Number of TNF⁺IL6⁺ MNPs per MV (right). (B) Mean fluorescent intensity (MFI) of TNF in MV MNPs from representative mIgG or α-CD47 treated K/B.g7 mice (left), MFI of TNF in MNPs per sample (middle), and number of TNF⁺ MNPs per MV (right). (C) MFI of IL6 in MV MNPs from representative mIgG or α-CD47 treated K/B.g7 mice (left), MFI of IL6⁺ in MNPs per sample (middle), and number of IL6⁺ MNPs per MV (right). (A-C) MNPs identified via flow cytometry as live, single, CD45.2⁺Ly6G⁻B220⁻CD3ε⁻SIRPα⁺ cells. Flow cytometric data were analyzed using a two-tailed student’s t-test. Points indicate averages for individual mice; red bars indicate mean ± SD.

We also evaluated the effect of in vivo CD47 blockade on circulating cytokine and chemokine levels. We found that the preventative CD47 blockade reduced the serum level of IL-6 and CXCL1 (Supplemental Table 1). Serum levels of TNF were low and not affected by CD47 blockade.

Taken together, our findings support a working model in which the accumulation of CD47-expressing apoptotic cells alongside phagocytes “frustrated” by the interaction of CD47 with SIRPα cooperate to promote macrophage cytokine production leading to valve inflammation and eventual fibrosis. CD47 blockade increases efferocytosis of apoptotic cells and reduces macrophage production of the pathogenic cytokines TNF and IL-6, culminating in reduced mitral valve inflammation.

Discussion

Efferocytosis of apoptotic cells occurs normally as part of developmental tissue remodeling, including formation of the cardiac valves (21). In the context of pro-inflammatory signals such as TNF and IL-6, apoptotic cells express CD47, a “don’t-eat-me” signal for SIRPα-expressing macrophages (7). We report increased expression of CD47 and increased presence of apoptotic cells in both murine and human valvular carditis. Antibody blockade of CD47 reduced the severity of valvular carditis in our mouse model. Mechanistically, we found that CD47 blockade increased macrophage uptake of apoptotic cells, reduced macrophage production of the pathogenic cytokines TNF and IL-6, and reduced TNF- and IL-6-dependent VCAM-1 expression. Collectively, these findings suggest an amplification loop is established that perpetuates chronic valve inflammation: macrophage-derived inflammatory cytokines act on valve resident cells to induce apoptosis as well as expression of CD47; the CD47:SIRPα interaction impairs efferocytosis and also potentiates macrophage cytokine production.

Our prior studies have demonstrated that macrophage stimulation through activating FcRs promotes IL-6 and TNF production in this model (16, 18). It has been previously shown that CD47 engagement of the inhibitory SIRPα-signaling pathway interferes with activating FcR-mediated signaling, reducing macrophage phagocytosis of opsonized erythrocytes and that this inhibitory pathway depends on the intracellular tyrosine phosphatase SHP-1 (25). Based on this prior report, we speculate that a similar interaction of FcR-activating and SIRPα-inhibitory signaling pathways is at play in the K/B.g7 model, leading to reduced macrophage uptake of CD47-expressing apoptotic cells. However, the situation is more complex, since macrophage cytokine production is simultaneously elevated while the CD47:SIRPα interaction blocks efferocytosis, a scenario reminiscent of “frustrated phagocytosis” (26). Further, some studies utilizing intact anti-CD47 blocking antibodies are complicated by the possibility that the Fc portion of the blocking antibodies can engage activating FcRs; while it is possible that this mechanism could account for a portion the increased uptake of apoptotic cells we observed following CD47 blockade, it would not readily account for the reduction in macrophage TNF and IL-6 expression. Interestingly, CD47 expression on endothelial cells has been demonstrated to play a role in leukocyte transendothelial migration, particularly of T cells (27). It remains possible that this CD47-dependent pathway is also at play in the K/B.g7 model, although T cells are not a dominant valve-infiltrating cell type (16). We therefore favor the interpretation that CD47 blockade relieves SIRPα-mediated inhibitory signaling, allowing effective macrophage efferocytosis of apoptotic cells and resolution of “frustrated” efferocytosis.

Our analysis revealed that multiple cell types undergo apoptosis in this model (Figure 1). In addition to these apoptotic cells promoting macrophage cytokine production, we further postulate that unengulfed apoptotic cells/debris might serve as a source of additional autoantigen for recognition by anti-GPI autoantibodies, providing additional activating signals to local macrophages via activating FcR signaling (16, 18).

Targeting the CD47:SIRPα interaction with either monoclonal antibodies or nanoparticles reduced diseased severity in experimental models of atherosclerosis (7, 28). Our findings demonstrate that targeting CD47 can be therapeutically effective in other forms of chronic cardiovascular inflammation, specifically autoimmune valvular carditis. Based on observations in our mouse model and its multiple correlates to human RHD, we suggest that targeting the CD47:SIRPα interaction could be effective for patients with chronic autoimmune cardiac valve inflammation and fibrosis.

Key Points.

• CD47-expressing apoptotic cells accumulate in autoimmune valvular carditis.

• CD47 blockade promotes efferocytosis and reduces inflammatory cytokine production.

• CD47 blockade reduces the severity of autoimmune valvular carditis in mice.

Abbreviations:

GPI

glucose-6-phosphate isomerase

RHD

rheumatic heart disease

SIRPα

signal recognition protein alpha

Syk

spleen tyrosine kinase