Key Points
SIRPα deficiency enables various inflammatory stimuli to confer sHLH/CSS in mice.
Macrophage depletion, but not IFN-γ neutralization, precludes sHLH in SIRPα−/− mice.
SIRPα negatively regulates TLR9 signaling by inhibiting Erk1/2 and p38 activation.
Visual Abstract
Abstract
A hallmark of secondary hemophagocytic lymphohistiocytosis (sHLH), a severe form of cytokine storm syndrome, is the emergence of overactivated macrophages that engulf healthy host blood cells (i.e., hemophagocytosis) and contribute to the dysregulated inflammation-driven pathology. In this study, we show that depleting SIRPα (SIRPα−/−) in mice during TLR9-driven inflammation exacerbates and accelerates the onset of fulminant sHLH, in which systemic hemophagocytosis, hypercytokinemia, consumptive cytopenias, hyperferritinemia, and other hemophagocytic lymphohistiocytosis hallmarks were apparent. In contrast, mice expressing SIRPα, including those deficient of the SIRPα ligand CD47 (CD47−/−), do not phenocopy SIRPα deficiency and fail to fully develop sHLH, albeit TLR9-inflamed wild-type and CD47−/− mice exhibited hemophagocytosis, anemia, and splenomegaly. Although IFN-γ is largely considered a driver of hemophagocytic lymphohistiocytosis pathology, IFN-γ neutralization did not preclude the precipitation of sHLH in TLR9-inflamed SIRPα−/− mice, whereas macrophage depletion attenuated sHLH in SIRPα−/− mice. Mechanistic studies confirmed that SIRPα not only restrains macrophages from acquiring a hemophagocytic phenotype but also tempers their proinflammatory cytokine and ferritin secretion by negatively regulating Erk1/2 and p38 activation downstream of TLR9 signaling. In addition to TLR9 agonists, TLR2, TLR3, or TLR4 agonists, as well as TNF-α, IL-6, or IL-17A, but not IFN-γ, similarly induced sHLH in SIRPα−/− mice but not SIRPα+ mice. Collectively, our study suggests that SIRPα plays a previously unappreciated role in sHLH/cytokine storm syndrome pathogenesis by preventing macrophages from becoming both hemophagocytic and hyperactivated under proinflammation.
This article is featured in Top Reads, p.2555
Introduction
Cytokine storm syndromes (CSS) such as secondary hemophagocytic lymphohistiocytosis (HLH) (sHLH) are life-threatening disorders associated with bacterial and viral infections and have recently been considered a cause for mortality in persons infected with SARS-CoV-2 (1–4). Also known as macrophage activation syndrome (5, 6), sHLH is associated with hyperactivated macrophages that adopt a hemophagocytic phenotype and thus aberrantly engulf other healthy hematopoietic-lineage cells such as RBCs, leukocytes, and platelets, often leading to consumptive cytopenias (i.e., anemia, leukopenia, and thrombocytopenia) (7–9). The underlying mechanism driving macrophages to become hemophagocytic under inflammatory conditions, however, has yet to be clearly determined (10).
Although an ensemble of proteins regulates macrophage phagocytosis, the inhibitory receptor that is well known to prevent phagocytosis of healthy host cells is SIRPα (11, 12). The SIRPα–CD47 axis serves as the quintessential “don’t-eat-me” signal to prevent unwanted phagocytosis of healthy host cells (13–16). As an inhibitory receptor highly expressed on myeloid leukocytes including macrophages, monocytes (MC), dendritic cells (DC), and neutrophils (PMN), SIRPα is critical to prohibiting phagocytosis of CD47-expressing host cells such as RBC, leukocytes, and platelets (17, 18). Despite the importance of this safety mechanism, the absence of SIRPα or CD47 alone does not confer an autoimmune phenotype as mice devoid of SIRPα (SIRPα−/−) or CD47 (CD47−/−) are healthy like their wild-type (WT) littermates (8, 19). Providing insight into this paradoxical phenomenon, our prior study revealed that macrophages will not phagocytose healthy host cells unless two criteria are simultaneously met: 1) SIRPα–CD47 inhibition must be removed, and 2) proinflammatory activation signaling must be present, presumably to activate the uncharacterized macrophage phagocytic receptor recognizing healthy host cells (8). Although this bipartite “fail-safe” mechanism has yet to be fully understood, we have shown that various TLR agonists or certain proinflammatory cytokines such as IL-1β, IL-6, TNF-α, or IL-17A, but not IFN-γ, are capable of driving the required activation signaling (8, 20).
Aside from phagocytosing healthy blood cells, overactivated macrophages have been reported to contribute to other cardinal features commonly presenting in fulminant sHLH including hyperferritinemia, hypertriglyceridemia, and hypercytokinemia, the latter being the putative driver of multiorgan failure and death (5, 8, 21–25). Indeed, previous reports have shown that dysregulation of proinflammatory macrophage activation can contribute to HLH/CSS pathogenesis in various models of sterile and nonsterile inflammation (9, 26, 27). Along these lines, SIRPα not only canonically inhibits macrophage phagocytosis but has become increasingly appreciated to also negatively regulate macrophage-driven proinflammation, with inflammatory diseases such as colitis, peritonitis, and diabetic nephropathy being exacerbated under SIRPα deficiency (8, 28–30). Thus, we hypothesized that SIRPα likely prevents the overactivation of macrophages to the point of becoming hemophagocytic under proinflammation and potentially serves as a critical barrier in the macrophage-intrinsic contribution to sHLH/CSS pathogenesis.
Materials and Methods
Mice
Mice were 8–11 wk of age and sex matched for all studies. WT (C57BL/6J) and CD47−/− (B6.129-CD47tm1Fpl/J) mice were purchased from The Jackson Laboratory. SIRPα−/− mice were established as previously described (8). All mice were housed in a specific pathogen-free facility. All animal experiments and procedures for handling were approved by the Institutional Animal Care and Use Committee of Georgia State University.
Treatments
In CpG experiments, WT, SIRPα−/−, or CD47−/−−/− mice were injected with zymosan A (500 μg/mouse, i.p.; Sigma), poly I:C (10 mg/kg, i.p.; InvivoGen), LPS (0.25 mg/kg, i.p.; Escherichia coli O111:B4; Sigma), rTNF-α (10 μg/kg, i.v.; PeproTech), rIL-6 (10 μg/kg, i.v.; PeproTech), rIL-17A (10 μg/kg, i.v.; PeproTech), or rIFN-γ (10 μg/kg, i.v.; PeproTech) every other day for a total of three injections and were euthanized 48 h thereafter for analyses (8).
Hematology
Disease progression was monitored daily by changes in hemoglobin (Hgb), which was assessed by serial sampling of blood (30–40 μl) via the saphenous vein. Hgb concentration was determined by adding 25 μl of whole blood to diluted Drabkin reagent (Sigma) and then measuring the spectral absorbance of hemiglobincyanide at 540 nm. Percentage change in Hgb was determined by 100 − (measured Hgb/initial Hgb) × 100. To analyze peripheral blood at the treatment end point, mice were euthanized, and blood was extracted by cardiac puncture into heparinized tubes. One hundred microliters of whole blood was then used to isolate platelets (31), which were then counted on a Cellometer Auto 2000 (Nexcelom Bioscience). The remaining whole blood was then centrifuged (1000 × g) to isolate the plasma for later quantification of cytokines and other biomarkers. The buffy coat and RBC pellet were resuspended in lysis buffer (150 mM ammonium chloride, 10 mM potassium bicarbonate, and 0.1 mM EDTA) for RBC lysis, and the remaining leukocytes were then counted on a Cellometer Auto 2000.
Histology and microscopy
Freshly isolated spleens and livers were embedded in OCT compound, flash frozen in liquid nitrogen, and then stored at −80°C. Tissues were cryostat sectioned at 6–10 μm and then immediately fixed in 4% paraformaldehyde. H&E stains were performed for histological analysis of hemophagocytosis in bone marrow smears, spleens, and livers. For adoptive transfer experiments, mice were first i.v. injected with ∼109 CFSE-stained RBC, and after 1 h, mice were i.p. injected with CpG (50 μg) or PBS (8). After 12 h, mice were euthanized, and their spleens removed. In some experiments, spleens were digested, and splenocytes were isolated for flow cytometry to quantify the percentage of hemophagocytic (CFSE+2O2
Macrophage phagocytosis assay
Bone marrow–derived macrophages (BMM) were generated as previously described (8, 34). To assess phagocytic activity, BMM were treated with PBS, CpG (1 μg/ml), poly I:C (100 ng/ml), LPS (20 ng/ml), TNF-α (20 ng/ml), or IFN-γ (20 ng/ml) for 12 h and then were washed and incubated with peripheral blood cells (PBC) for 30 m at 37°C. BMM–PBC cocultures were then washed, extracellular RBC were lysed, and then cells were visualized by microscopy. For IF staining, PBC were labeled with CFSE prior to coculture with BMM. BMM were then labeled with PE-conjugated anti-F4/80 and fixed with 4% paraformaldehyde. After fixation, nuclei were stained with DAPI, and cells were visualized by fluorescent microscopy.
Cytokine and biomarker quantification
Soluble mCD47.ex
The alkaline phosphatase tag2 plasmid containing the mCD47.ex and alkaline phosphatase was a generous gift of V. Narayanan (University of Pittsburgh School of Medicine) (35). After transfecting COS cells, recombinant mCD47.ex fusion proteins were produced, affinity purified, and stored in PBS as previously described (18, 35–38). Prior to use, the binding capacity of mCD47.ex was confirmed by binding to murine SIRPα extracellular domain fusion protein as well as adhesion to macrophages in a CD47-SIRPα–dependent manner.
SDS-PAGE and immunoblot
Statistical analysis
All graphs and statistical analyses were generated and performed using Prism 6.0 (GraphPad Software). Statistical significance was calculated using Student t test (k = 2) or ANOVA (one or two way) (k > 2). For post hoc analyses, Dunn or Tukey honestly significant difference (HSD) test was used to determine statistical significance among multiple comparisons, with an experiment-wise error rate of 0.05. When p < 0.05, samples were considered statistically significant. At least three independent experiments were performed for each set of data, which were represented as individual values, mean ± SEM, or both.
Results
SIRPα deficiency exacerbates and accelerates the onset of TLR9-induced sHLH
As a TLR9 agonist, CpG is both a relevant TLR stimulus in sHLH and a viable trigger for activating SIRPα−/− phagocytes to engulf healthy host cells (8, 39). To examine the role of SIRPα in TLR9-induced sHLH, we repeatedly injected WT or SIRPα−/− mice with CpG. The development of sHLH was monitored by measuring changes in Hgb. As shown (Fig. 1A, left panel), whereas WT mice did not exhibit anemia (i.e., ≥20% reduction in Hgb), until 24 h after the third injection of CpG, SIRPα−/− mice experienced a ∼25% drop in Hgb after merely one dose of CpG. With every additional dose of CpG, SIRPα−/− mice became increasingly more anemic and decreased their Hgb in excess of 60% within 48 h after three injections (day 6), which led us to terminate the study (Fig. 1A). By day 6, CpG-treated WT mice were only slightly anemic, whereas SIRPα−/− mice had severe pancytopenia (Fig. 1A, 1B). Assaying other pathological features of sHLH, we found that plasma ferritin, triglycerides, and sCD25/IL-2Rα were highly elevated in CpG-treated SIRPα−/− mice, whereas WT mice only had a slight increase in sCD25 (Fig. 1C–E) (1). Screening resected organs found that WT mice had normal livers and only moderate splenomegaly (Fig. 1F, 1G). However, CpG-treated SIRPα−/− mice had severe hepatosplenomegaly, accounting for ≥10% of their body weight. Histological examination of CpG-treated spleens revealed that the splenic white pulp architecture remained uniform in WT mice but was greatly disrupted in SIRPα−/− mice (Supplemental Fig. 2A). Indeed, splenic leukocyte expansion was commensurate to the severity of splenomegaly, with CpG-treated SIRPα−/− mice having 2–3-fold more CD45+ splenocytes than WT mice (Supplemental Fig. 2C). Healthy SIRPα−/− mice had slightly lower frequencies of DC and CD4 T cells (Fig. 1H, Supplemental Fig. 2C), corroborating previous findings (8, 40, 41). After CpG treatment, SIRPα−/− mice had lower splenic frequencies of B cells and CD4 T cells than WT mice; however, adjusting for cell count mitigated these differences as SIRPα−/− mice actually had 20% more B cells and a similar quantity of CD4 cells as WT mice (Supplemental Fig. 2C). Whereas neither the frequency nor quantity of splenic myeloid leukocytes increased in CpG-treated WT mice, PMN, MC, and RPM had increased 2–3-fold in SIRPα−/− mice (Fig. 1H). Thus, myeloid cell and CD8 T cell expansion appeared to account for the reduced B cell frequency in CpG-treated SIRPα−/− spleens (Supplemental Fig. 2E).
SIRPα deficiency exacerbates and accelerates the onset of TLR9-induced sHLH. (A–I) WT or SIRPα−/− mice were challenged with three i.p. injections of CpG-B (oligodeoxynucleotide 1864; 50 μg/mouse; n = 9/group) or PBS (100 μl/mouse; n = 5/group) on days 0, 2, and 4 and were euthanized on day 6 for analyses. (A) Disease severity was monitored by measuring daily changes in Hgb (left). (A and B) Following euthanization on day 6, anemia [(A), right], leukopenia [(B), left], and thrombocytopenia [(B), right] were quantified by peripheral blood analyses; (C–F) additional blood markers of plasma ferritin (C), triglycerides (D), and sCD25 (E) were determined by ELISA. (F and G) Hepatosplenomegaly. Livers (F) and spleens (G) were weighed and normalized to their respective total body weight. Representative image of splenomegaly [(G), bottom]. (H) Myeloid population frequencies (left) and counts (right) among CD45+ splenocytes were determined by flow cytometry: CD11b−F4/80+ RPM, CD11b+Ly6G−Ly6C+ MC, CD11b+Ly6Ghi PMN, and CD11c+ DC (Supplemental Fig. 1: gating strategy). (I) Concentrations of plasma cytokines (IL-1β, IL-6, IL-12, TNF-α, MCP-1, and GM-CSF) from WT or SIRPα−/− mice 12 h after receiving an injection of PBS or CpG. The data represent 12 independent experiments, and all data were measured in triplicate. Each symbol represents an individual mouse, with the mean ± SEM also shown. Two-way ANOVA and Tukey HSD post hoc analyses were used to determine statistical significance among multiple comparisons. ns indicates p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Being a CSS, sHLH is typified by elevated proinflammatory cytokines, with overactivated macrophages being a major source (42–45). As an absence of SIRPα worsens disease conditions (8, 16, 28–30, 46), we assessed whether SIRPα deficiency would exacerbate CpG-induced hypercytokinemia. Healthy WT and SIRPα−/− mice mostly exhibited similar cytokine levels; however, under TLR9 stimulation, SIRPα−/− mice had 2–4-fold higher levels of IL-1β, IL-6, IL-12, IL-18, and TNF-α than WT mice (Fig. 1I, Supplemental Fig. 2G). The exacerbated phenotype in CpG-treated SIRPα−/− mice was not due to a stunted anti-inflammatory response as both mice produced comparable amounts of IL-10 after CpG treatment (Supplemental Fig. 2G). Next, we assayed cytokines associated with granulopoiesis and myeloid leukocyte chemotaxis to determine what led to the expanded myeloid population in CpG-treated SIRPα−/− spleens (47–49). Similar to IL-10, GM-CSF was low in SIRPα−/− mice prior to TLR9 signaling, but after CpG, circulating GM-CSF significantly increased. The chemoattractant MCP-1 (CCL2) was also produced almost 3-fold greater in SIRPα−/− mice than in WT mice responding to TLR9 agonism. Collectively, these data suggest that an absence of SIRPα enhances both the kinetics and severity of sHLH onset induced by TLR9-driven inflammation (9, 50).
Absence of SIRPα during TLR9 stimulation induces hemophagocytic leukocytes
Our previous study showed that acute inflammation in SIRPα−/− mice led to erythrophagocytosis, consumptive anemia, and splenomegaly (8); thus, we next assessed the extent of hemophagocytosis within the spleen following TLR9 stimulation. To quantify splenic hemophagocytes, splenic leukocytes (CD45+) were intracellularly stained for the RBC-associated Ag Ter-119 (7, 32, 33, 51). Regardless of genotype, ∼10–20% of RPM were Ter-119+ under healthy conditions (Fig. 2A, 2B, Supplemental Fig. 3A), corroborating previous reports (7, 33). However, whereas the frequency of Ter-119+ RPM did not significantly increase in WT mice following CpG treatment, over 50% of RPM were Ter-119+ in the absence of SIRPα. Because CpG-treated SIRPα−/− spleens were highly infiltrated with other myeloid leukocytes that innately express SIRPα, we also assessed the frequency of Ter-119+ cells among MC, PMN, and DC (32). Without inflammatory stimulation, no differences in Ter-119+ MC, PMN, and DC frequency were detected. After CpG treatment, however, all SIRPα−/− myeloid cells exhibited robust hemophagocytosis (Fig. 2A, 2B), whereas the quantity of Ter-119+ cells did not significantly increase in WT mice. On average, CpG-treated SIRPα−/− mice spleens comprised 18-fold more Ter-119+ leukocytes than CpG-treated WT mice (Fig. 2B). Indeed, H&E staining of spleen and liver tissue sections as well as bone marrow smears were insufficient to detect hemophagocytosis in CpG-treated WT mice (data not shown), whereas hemophagocytosis was apparent in tissues isolated from SIRPα−/− mice (Fig. 2C). To determine whether this difference in hemophagocytosis may occur after just one CpG injection, we adoptively transferred fluorescence-labeled (CFSE) RBC with CpG or PBS concurrently injected (8). In mice receiving CpG, 7% of RPM were CFSE+ in WT mice, whereas 40% were CFSE+ in SIRPα−/− mice (Fig. 2D). IF staining of spleen sections from CpG-treated mice supported our flow cytometry data (Fig. 2E, inset), with the frequency of CFSE+F4/80+ macrophages (yellow) being much more apparent in spleens of CpG-treated SIRPα−/− mice.
Absence of SIRPα during TLR9 stimulation induces hemophagocytic leukocytes. (A) Frequencies of hemophagocytic CD45+ splenocytes (F4/80+CD11b− RPM, CD11b+Ly6G−Ly6C+ MC, CD11b+Ly6Ghi PMN, and CD11c+ DC) were determined by intracellular staining for RBC (Ter-119+). Isotype controls are shown in Supplemental Fig. 3A. (B) Frequencies (left), individual counts (middle), and total counts (right) of hemophagocytes from PBS- or CpG-treated mice (n = 5/group). (C) Representative images of hemophagocytes within the bone marrow, spleen, and liver of CpG-treated SIRPα−/− mice as determined by H&E staining (scale bar, 10 μm) (n = 3/group). (D) Histogram frequency of CD45+CD11b−F4/80+ RPM with intracellular RBC (CFSE+). (E) Representative image of IF-stained spleen tissue sections after RBC (109) transfer in CpG-treated WT and SIRPα−/− mice (scale bar, 100 μm). The data represent eight independent experiments, and each symbol represents an individual mouse, with the mean ± SEM also shown. Two-way ANOVA and Tukey HSD post hoc analyses were used to determine statistical significance among multiple comparisons. ns indicates p > 0.05, ****p < 0.0001.
CD47 deficiency does not phenocopy SIRPα deficiency under TLR9 agonism
Serving as a ligand for SIRPα, CD47 is a critical self-recognition molecule expressed on healthy host cells, which are normally not phagocytosed (19). One study previously reported that serum from HLH patients can downregulate CD47 on hematopoietic stem cells (HSC), which led macrophages to phagocytose HLH serum-treated HSC in vitro. However, the question remains: do CpG-treated CD47−/− mice similarly develop fulminant sHLH as in SIRPα−/− mice? Following CpG treatment, Hgb sharply dropped in CD47−/− mice, which largely mimicked the kinetics and severity in CpG-treated SIRPα−/− mice (Fig. 3A). However, unlike CpG-treated SIRPα−/− mice, CD47−/− mice did not develop leukopenia or thrombocytopenia (Fig. 3B). CpG-treated CD47−/− mice had slightly elevated ferritin, albeit not comparably as severe in SIRPα−/− mice (Fig. 3C). Furthermore, both plasma triglycerides and sCD25 were not highly elevated in CpG-treated CD47−/− mice (Fig. 3D, 3E). As anticipated, CD47−/− mice developed splenomegaly after TLR9 stimulation (Fig. 3F, Supplemental Fig. 2B) (8); however, unlike SIRPα deficiency, an absence of CD47 only led to moderate expansion of RPM and a slight increase in splenic infiltration of PMN (Fig. 3G, Supplemental Fig. 2D, 2F). Despite exhibiting a noticeable increase in Ter-119+ RPM, CpG-treated CD47−/− spleens comprised exceptionally less hemophagocytes than SIRPα−/− spleens, which had nearly 5-fold more Ter-119+ cells than CD47−/− mice (Fig. 3H, Supplemental Fig. 2H, 2I). Assessing plasma cytokines found that CD47−/− mice were appreciably less inflamed than SIRPα−/− mice (Fig. 3I). By calculating H-scores (1), we found that CpG-treated CD47−/− and WT exhibited similar degrees of severity, and both had H-scores of 58, equating to a <1% probability of having HLH. In stark contrast, CpG-treated SIRPα−/− mice had an H-score of 186, indicating a 70–80% probability of having HLH. These data suggest that SIRPα negatively regulates inflammation in a CD47-independent manner and that although CD47 deficiency is sufficient to develop hemophagocytosis, anemia, and splenomegaly, it does not fully mimic SIRPα deficiency with respect to fully developing sHLH.
CD47 deficiency does not phenocopy SIRPα deficiency under TLR9 agonism. (A–I) SIRPα−/− or CD47−/− mice were injected three times with CpG (n = 5/group) or PBS (n = 5/group) on days 0, 2, and 4 and euthanized on day 6 for analyses. (A) Daily changes in Hgb (left). Following euthanization on day 6, anemia [(A), right], leukopenia [(B), left], and thrombocytopenia [(B), right] were quantified, and additional blood markers of plasma ferritin (C), triglycerides (D), and sCD25 (E) were determined. (F–H) Spleens were isolated to determine splenomegaly (F), the frequency of myeloid leukocytes (G), and the frequency [(H), left] and count of splenic hemophagocytes [(H), right]. (I) Plasma cytokines (MCP-1, GM-CSF, IL-6, and TNF-α) assayed 12 h following a single CpG or PBS injection. The data represent nine independent experiments, and all data were measured in triplicate. Each symbol represents an individual mouse, with the mean ± SEM also shown. Two-way ANOVA and Tukey HSD post hoc analyses were used to determine statistical significance among multiple comparisons. ns indicates p > 0.05, *p < 0.05, ***p < 0.001, ****p < 0.0001.
Macrophage depletion but not IFN-γ neutralization ameliorates TLR9-induced sHLH in SIRPα−/− mice
Hypercytokinemia is largely considered to precede multiorgan failure and death in CSS/HLH. Accordingly, current efforts to treat HLH/CSS primarily revolve around neutralization of cytokines considered putative drivers of the immunopathology. IFN-γ is one such cytokine that is considered critical to manifesting immunopathological features of HLH (25, 39, 50), and indeed, previous studies have shown that neutralization of IFN-γ may reverse sHLH symptoms in the TLR9–sHLH mouse model, even when exacerbated by coadministering antagonistic IL-10R Abs (39, 50, 52, 53). Thus, we assessed whether IFN-γ neutralization could similarly prove to be therapeutic in CpG-treated SIRPα−/− mice by coadministering anti–IFN-γ Ab (XMG1.2) or isotype control (Fig. 4). To our surprise, IFN-γ neutralization was largely ineffective in preventing sHLH onset in SIRPα−/− mice, suggesting sHLH pathogenesis under SIRPα deficiency is not dependent upon IFN-γ.
Macrophage depletion but not IFN-γ neutralization precludes TLR9-induced sHLH in SIRPα−/− mice. SIRPα−/− mice were treated with PBS (n = 5) or CpG coadministered with IgG isotype control (n = 5) or anti–IFN-γ (XMG1.2; n = 5) on days 0, 2, and 4 and euthanized on day 6 for analyses. To deplete macrophages, EL (n = 5) or CL2MDP (n = 5) were administered prior (day −2 and −1) to CpG treatment. (A) Daily changes in Hgb (left). Following euthanization on day 6 (A–E), anemia [(A), right], leukopenia [(B), left], and thrombocytopenia [(B), right] were quantified, and additional blood markers of plasma ferritin (C), triglycerides (D), and sCD25 (E) were determined. (F–I) Spleens were isolated to determine splenomegaly (F and G), the frequency of myeloid leukocytes (G), and the frequency [(H), left] and the count [(H), middle and left] of hemophagocytes. (J) Plasma cytokines (IL-1β, IL-6, IL-12, TNF-α, and GM-CSF) assayed 12 h following a single injection of PBS, CpG + IgG isotype control, CpG + anti–IFN-γ, or CpG alone after EL or CL2MDP treatment. The data represent 13 independent experiments, and all data were measured in triplicate. Each symbol represents an individual mouse, with the mean ± SEM also shown. One-way ANOVA and Dunn post hoc analyses were used to determine statistical significance among multiple comparisons. ns indicates p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Our prior study implicated macrophages as drivers of anemia in inflamed SIRPα−/− mice (8); thus, we assessed the therapeutic efficacy of macrophage depletion by administering EL or clodrosome-containing liposomes (CL2MDP) prior to (day −2 and −1) CpG treatment (26, 27, 54–57). In contrast to IFN-γ neutralization, macrophage depletion by CL2MDP precluded sHLH in CpG-treated SIRPα−/− mice. CL2MDP prevented the rapid onset of severe pancytopenia, with CpG-treated SIRPα−/− mice only exhibiting a 10% drop in Hgb by day 6 (Fig. 4A, 4B). Furthermore, macrophage depletion drastically reduced plasma ferritin and triglyceride levels and slightly lowered sCD25 levels as well (Fig. 4C–E). Assessing resected spleens, CL2MDP effectively prevented splenomegaly and also led to a significant reduction in myeloid splenocyte frequency and count (Fig. 4F–H, Supplemental Fig. 3B). Although CL2MDP did not significantly affect the frequency of Ter-119+ splenocytes, adjusting for cell count revealed a marked reduction in the quantity of splenic hemophagocytes (Fig. 4I). Notably, CL2MDP largely abated hypercytokinemia, albeit GM-CSF remained moderately elevated (Fig. 4J). These data suggest that phagocytes, chiefly macrophages, are primarily responsible for TLR9-driven sHLH pathology in SIRPα−/− mice.
SIRPα negatively regulates TLR9-driven Erk1/2 and p38 activation to inhibit macrophage function
Given CL2DMP treatment abrogated sHLH onset in CpG-treated SIRPα−/− mice, we sought to determine whether SIRPα directly affects TLR9 signaling in macrophages. As anticipated, SIRPα inhibited CpG-treated WT and CD47−/− BMM from phagocytosing WT (CD47+) PBC (Fig. 5A). Indeed, fluorescent-labeled CpG-activated WT or CD47−/− BMM (red) failed to phagocytose CFSE-positive PBC (green), whereas similarly treated SIRPα−/− BMM were highly phagocytic after CpG treatment and engulfed many PBC, including RBC and nucleated cells (arrows). Aside from being more hemophagocytic, SIRPα−/− BMM were adept at secreting proinflammatory cytokines (Fig. 5B). To determine the effect of SIRPα on BMM cytokine production, WT, CD47−/−, and SIRPα−/− BMM were treated with CpG in the presence or absence of CD47 ligation via mCD47.ex (18, 35–38). In the absence of mCD47.ex, CpG-treated WT and CD47−/− BMM produced on average 2-fold less IL-1β, IL-6, IL-12, and TNF-α than SIRPα−/− BMM (Fig. 5B). Conversely, SIRPα−/− BMM produced markedly less IL-10 than WT or CD47−/− BMM in response to TLR9 signaling. Given that previous studies have shown macrophages are a major source of extracellular ferritin (24, 58, 59), BMM-conditioned culture medium was also assayed for extracellular ferritin (Fig. 5C). Indeed, SIRPα−/− BMM secreted significantly more ferritin after CpG treatment than WT or CD47−/− BMM. Furthermore, the disparity between WT or CD47−/− BMM and SIRPα−/− BMM insofar as their capacities to secrete proinflammatory cytokines and ferritin was magnified in the presence of CD47 ligation (+mCD47.ex) (Fig. 5B, 5C). Along these lines, CpG-treated WT and CD47−/− BMM produced even more anti-inflammatory cytokine IL-10 in the presence of mCD47.ex, whereas SIRPα−/− BMM production of IL-10 remained limited under TLR9 signaling. Given that CD47–SIRPα signaling appeared to impact BMM cytokine production in response to TLR9 stimulation, we examined potential differences in macrophage TLR9 signaling in the presence or absence of CD47–SIRPα signaling (Fig. 5D). The extent to which Erk1/2 (p44/42) and p38 were phosphorylated was mostly similar among BMM after CpG treatment alone. However, in the presence of CD47 ligation (+mCD47.ex) and thus strong SIRPα signaling, CpG-treated WT, and CD47−/− BMM exhibited a level of phosphorylation of Erk1/2 and p38 that paralleled that of nontreated BMM, whereas SIRPα−/− BMM were unaffected and maintained MAPK activation. Together, these data suggest that SIRPα tempers TLR9 signaling in macrophages by inhibiting MAPK.
SIRPα negatively regulates TLR9-driven Erk1/2 and p38 activation to inhibit macrophage function. (A) BMM were generated from WT, CD47−/−, or SIRPα−/− mice, treated with CpG (1 μg/ml) and then cocultured with PBC isolated from healthy WT mice. Representative microscopy images of IF-stained BMM (PE-F4/80; red) and PBC (CFSE; green), with nuclei stained by DAPI (blue). The data represent three independent experiments, and each phagocytosis assay was performed in triplicate. (B and C) WT, CD47−/−, and SIRPα−/− BMM were treated with CpG (1 μg/ml) in the presence or absence of CD47 ligation (±mCD47.ex) for 12 h, and cell-free medium was then assayed by ELISA to quantify IL-1β, IL-6, IL-10, IL-12, TNF-α (B), and ferritin (C) secretion. The data represent four independent experiments, and each experimental group was performed in triplicate. The data are presented as mean + SEM. Two-way ANOVA and Tukey HSD post hoc analyses were used to determine statistical significance among multiple comparisons. (D) Representative immunoblot analyses of SIRPα, p38 phosphorylation (p-p38), Erk1/2 phosphorylation (p-p44/42), and β-actin protein abundance in WT, CD47−/−, and SIRPα−/− BMM that were nontreated, treated with mCD47.ex alone, or treated with CpG (1 μg/ml) in the presence or absence of CD47 ligation (±mCD47.ex) for 30 min. All immunoblots are representative of five independent experiments. Densitometric analysis was used to determine the relative change in phosphorylated p44/42 and p38 by normalizing against β-actin. ns indicates p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Activation of SIRPα−/− macrophages by various inflammatory factors confers sHLH-like disease
Given other TLR agonists and certain cytokines fulfill the requisite activation signaling for phagocytosis of healthy host cells (8), we treated mice with zymosan A (TLR2), poly I:C (TLR3), LPS (TLR4), TNF-α, IL-6, or IL-17A. One study has shown that IFN-γ alone induces a partial HLH phenotype in mice (7); thus, we also assessed IFN-γ–treated mice. Following zymosan, LPS, or IFN-γ treatment, WT mice developed mild anemia (Supplemental Fig. 4A). LPS treatment also induced splenomegaly and elevated ferritin (Supplemental Fig. 4D, 4E). Otherwise, all treatments failed to develop sHLH symptoms in WT mice. In contrast, treating SIRPα−/− mice with any inflammatory stimuli, except IFN-γ, led to an sHLH phenotype as that induced by CpG (Fig. 6). All hemophagocytosis-activating factors conferred a moderate to severe drop in Hgb (Fig. 6A). Whereas all treatments induced severe thrombocytopenia, poly I:C, TNF-α, and IL-6 only induced mild/moderate leukopenia (Fig. 6B, 6C). All treatments, except IFN-γ, led to severe splenomegaly (Fig. 6D), hyperferritinemia (Fig. 6E), and hypertriglyceridemia (Fig. 6F). Immunohistochemical staining against F4/80 in spleen sections and in vitro phagocytosis assays with poly I:C, LPS, or TNF-α, but not IFN-γ, demonstrated the exceptional capacity of activated SIRPα−/− phagocytes to uptake RBC (Fig. 6G). Quantification of plasma cytokines in LPS-treated SIRPα−/− mice again demonstrated markedly worse hypercytokinemia under SIRPα deficiency (Fig. 6H). H-scores were calculated and summarized in Fig. 7. Collectively, these data indicate the importance of SIRPα in not only suppressing a hemophagocytic phenotype but also preventing the development of sHLH/CSS under various inflammatory conditions.
Activation of SIRPα−/− macrophages by various inflammatory factors confers sHLH-like disease. SIRPα−/− mice (n = 5 mice per treatment) were injected with PBS, zymosan A (500 μg/mouse, i.p.), poly I:C (10 mg/kg, i.p.), LPS (0.25 mg/kg, i.p.), rTNF-α (10 μg/kg, i.v.), rIL-6 (10 μg/kg, i.v.), rIL-17A (10 μg/kg, i.v.), or rIFN-γ (10 μg/kg, i.v.) on days 0, 2, and 4 and were then euthanized on day 6 for analyses. Peripheral blood markers for HLH were assessed: anemia (A), leukopenia (B), thrombocytopenia (C), hyperferritinemia (E), and hypertriglyceridemia (F). (D) Spleens were also excised and weighed to determine splenomegaly. Representative images of SIRPα−/− BMM treated with poly I:C (100 ng/ml), LPS (20 ng/ml), TNF-α (20 ng/ml), or IFN-γ (20 ng/ml) for 12 h and then were incubated with RBC isolated from healthy mice [(G), top]. Representative images of immunohistochemical staining against F4/80 among spleens isolated from poly I:C–, LPS-, TNF-α–, or IFN-γ–treated SIRPα−/− mice [(G), bottom]. (H) Twelve hours after one injection of LPS, the concentration of IL-1β, IL-6, IL-10, TNF-α, and GM-CSF were quantified. The data represent 16 independent experiments. Each symbol represents an individual mouse. One-way ANOVA and Dunn post hoc analyses (A–F) or Student t test (H) were used to determined statistical significance among multiple comparisons. NS indicates p > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Summary of HLH phenotype in mice with varied treatments. Six or five hallmarks were assessed in CpG-treated or other inflammatory stimuli–treated mice, respectively, in accordance with the HLH 2004 diagnostic criteria. An H-score ≥169 accurately classifies 90% of individuals and corresponds to a sensitivity of 93% and specificity of 86% for HLH diagnosis. NS indicates p > 0.05, +p < 0.05, ++p < 0.01, +++p < 0.001, ++++p < 0.0001. ND, not determined.
Discussion
The appearance of aberrant histiocytes phagocytosing healthy host cells in sHLH is largely attributed to the hypercytokinemic milieu (60, 61). However, the mechanism by which inflammatory signaling overactivates macrophages and other myeloid leukocytes to the point of manifesting a hemophagocytic phenotype remains unclear. This study identifies SIRPα as a critical deterrent against the induction of hemophagocytic myeloid leukocytes under inflammation and also the development of sHLH/CSS. We found that activation of SIRPα−/− leukocytes by TLR agonists (TLR2, TLR3, TLR4, or TLR9) or proinflammatory cytokines (TNF-α, IL-6, or IL-17A but not IFN-γ) leads to severe pancytopenia and other hallmarks of sHLH (1, 6). Although HLH-associated cytokines downregulate CD47 on HSC, allowing macrophages to phagocytose HSC in vitro (62), we found that TLR9-driven inflammation in CD47−/− mice, unlike SIRPα−/− mice, does not induce sHLH but rather only leads to hemophagocytosis, anemia, and splenomegaly (Figs. 3, 7). Conceivably, the differences between SIRPα and CD47 in cellular expression and function underly the inability of CD47−/− mice to fully develop severe sHLH under TLR9 agonism as in SIRPα−/− mice (63, 64). sHLH symptoms associated with macrophage activation, such as hypercytokinemia, hypertriglyceridemia, and hyperferritinemia, were less severe in CpG-treated CD47−/− mice than in SIRPα−/− mice (21, 24, 25). This disparity likely manifests because of the differing roles of SIRPα and CD47 in regulating macrophage activation, as we and others have shown that SIRPα tempers the proinflammatory macrophage phenotype (Fig. 5) (8, 29). Collectively, these observations suggest that there are CD47-independent SIRPα regulatory mechanisms that remain to be clarified.
Whereas SIRPα canonically suppresses phagocytosis of healthy host cells, several studies suggest that SIRPα also negatively regulates proinflammation as SIRPα deficiency exacerbates disease conditions (8, 28–30). Similarly, SIRPα deficiency worsened and accelerated the onset of TLR9-driven sHLH (Figs. 1, 2). Comparable severity of the sHLH phenotype has been shown to also manifest in CpG-treated mice with intact SIRPα signaling if they are deficient in IL-18BP, which opposes IL-18 signaling (50). However, unlike other sHLH models, TLR9-challenged SIRPα−/− mice rapidly exhibited severe anemia, hemophagocytosis, and hypercytokinemia following the initial CpG injection, which likely points to the involvement of innate immunity in sHLH pathology therein. Wang et al. (9) showed that a specific sequence of pathogen sensing (i.e., viral [TLR3] to bacterial [TLR4] challenge) leads to metabolic dysregulation in macrophages and subsequently induces a highly lethal hyperinflammatory state resembling HLH secondary to endotoxic shock. Similarly, Mahajan et al. (26, 27) has shown that dysregulation of lipid signaling in macrophages can also confer macrophage-intrinsic sHLH/CSS pathology. In parallel, our data suggest that macrophages and likely other myeloid leukocytes also become dysregulated and hyperinflammatory in the absence of SIRPα-dependent inhibition (Figs. 4, 5), leading to severe and rapid onset of macrophage-intrinsic sHLH/CSS pathology. Indeed, activating SIRPα−/− macrophages with TLR agonists or proinflammatory cytokines led to exuberant hemophagocytosis and macrophage-intrinsic fulminant sHLH (Figs. 6, 7), whereas depleting macrophages prevented sHLH onset in TLR9-inflamed SIRPα−/− mice (Fig. 4). Mechanistic studies suggest that this phenotype was partially due to SIRPα inhibiting TLR9-driven macrophage activation by negatively regulating MAPK pathways (Fig. 5); thus, removing SIRPα appears to “prime” macrophages and facilitates their acquisition of a hyperinflammatory hemophagocytic state. Other studies have similarly shown that SIRPα regulates M1 and M2 macrophage polarization by modulating PI3K-Akt, MAPK, and NF-κB pathways (29, 46). However, given that CL2MDP has been shown to be nonspecific (55, 65–67), further investigation is required to ascertain a definitive role for macrophages over other phagocytes in HLH/CSS pathogenesis in SIRPα−/− mice.
The proinflammatory cytokine IFN-γ is considered a key molecule driving inflammation in HLH and is intimately associated with hemophagocytosis and anemia of inflammation (7, 52, 68–70). For example, two HLH patients had not developed hemophagocytosis owing to an IFN-γR deficiency (71). Indeed, IFN-γ administration alone is sufficient to induce hemophagocytic RPM in WT mice, leading to consumptive pancytopenia (7). To that end, researchers have endeavored to treat HLH by neutralizing IFN-γ, which has demonstrated exceptional efficacy in various settings (39, 50, 68). In contrast, IFN-γ neutralization was nontherapeutic in SIRPα−/− mice (Fig. 4), and IFN-γ treatment failed to induce any aspects of sHLH in SIRPα−/− mice (Figs. 6, 7), corroborating our previous study showing that although many proinflammatory cytokines and TLR agonists may drive SIRPα−/− macrophages to become hemophagocytic, IFN-γ has no such effect (8). Thus, future studies are necessary to determine the role of IFN-γ in driving macrophages to become hemophagocytic if not for providing the putative activation signaling (8). Supporting IFN-γ–independent mechanisms in HLH onset, Albeituni et al. (53) compared the therapeutic efficacies of anti–IFN-γ and ruxolitinib, a JAK1/2 inhibitor, in murine models of HLH and found that ruxolitinib was superior as it reduced PMN expansion and tissue infiltration (72, 73). In parallel, an absence of SIRPα under TLR9-driven inflammation greatly increased granulopoiesis and PMN tissue infiltration, which may partially underly the exceptionally severe sHLH phenotype in CpG-treated SIRPα−/− mice and the inefficacy of anti–IFN-γ (Fig. 5). Interestingly, CL2MDP depletion reduced circulating GM-CSF and also the frequency and number of splenic PMN in CpG-treated SIRPα−/− mice (Fig. 6). Furthermore, we show the mechanism by which macrophages and other myeloid leukocytes may become hemophagocytic and drive HLH-like disease under, and SIRPα deficiency is partially redundant as an array of TLR agonists and proinflammatory cytokines are capable of providing the activation signaling (Figs. 6, 7). This heterogenous capacity to drive myeloid leukocytes toward a hemophagocytic phenotype likely underscores why hemophagocytosis is neither specific nor sensitive to HLH or CSS in general and may also lend an explanation as to why CSS/HLH patients differentially respond to current cytokine neutralization therapies (1, 53, 74–78).
Although our studies reveal a link between SIRPα and the nascence of both hemophagocytes and HLH-like disease, a major question remains unanswered: do events preceding the onset of HLH/CSS comprise a phase during which SIRPα becomes downregulated or lost? Another puzzle is whether SIRPα can be downregulated under physiological or pathological conditions. Although the first question is currently unanswered and certainly demands further investigation, some studies have shown particular TLR-driven inflammation or disease conditions such as diabetic nephropathy lead to or are associated with a loss of SIRPα expression (8, 20, 30, 46). Conceivably, these conditions not only provide the predisposing condition (i.e., an absence of SIRPα) but also the necessary inflammatory activation (i.e., TLR or proinflammatory cytokine signaling) to drive leukocytes toward a hemophagocytic phenotype and potentially confer HLH/CSS-like disease. Akilesh et al. (33) recently showed that chronic TLR7 and TLR9 signaling reprograms myelopoiesis toward differentiating specialized MC, referred to as inflammatory hemophagocytes (iHPCs). iHPCs arose in aged mice with constitutively active TLR7 (TLR7.1) or, to a lesser degree, in mice injected with R848 (TLR7 agonist) or CpG every day for 13 or 5 d, respectively. However, these iHPCs appear to represent a separate subset of hemophagocytes following long-term myelopoietic reprogramming (79, 80), and given the stark differences in the kinetics and severity of hemophagocytosis and cytopenias, these iHPCs likely differ from hemophagocytes that arise in the absence of SIRPα. Nevertheless, our findings collectively demonstrate that SIRPα plays an indispensable role in preventing myeloid leukocytes from becoming hemophagocytic under inflammation and may also provide an explanation for why heterogenous inciting inflammatory factors can inevitably culminate in HLH/CSS pathology.
Disclosures
The authors have no financial conflicts of interest.
Acknowledgments
We thank the staff in the Georgia State University Animal Resources Program for assisting in experiments and animal care. The visual abstract was created using BioRender (http://www.BioRender.com).
Footnotes
This work was supported in part by the National Institutes of Health, a National Institute of Allergy and Infectious Diseases Grant (R01AI06839), a National Cancer Institute Grant (R21CA241271), a Georgia Research Alliance Venture Development Grant, a Biolocity Innovation and Commercialization Grant, a Molecular Basis of Disease Fellowship from Georgia State University (to K.K.), an Ahmed T. Abdelal Molecular Genetics and Biotechnology Fellowship from Georgia State University (to K.K.), and a Careers in Immunology Fellowship from the American Association of Immunologists (to Z.B.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
- BMM
- bone marrow–derived macrophage
- CL2MDP
- clodrosome-containing liposome
- CSS
- cytokine storm syndrome
- DC
- dendritic cell
- EL
- empty liposome
- Hgb
- hemoglobin
- HLH
- hemophagocytic lymphohistiocytosis
- HSC
- hematopoietic stem cell
- HSD
- honestly significant difference
- IF
- immunofluorescent
- iHPC
- inflammatory hemophagocyte
- MC
- monocyte
- mCD47.ex
- extracellular domain of murine CD47
- PBC
- peripheral blood cell
- PMN
- neutrophil
- RPM
- red pulp macrophage
- sCD25
- soluble CD25
- sHLH
- secondary HLH
- WT
- wild-type.
- Received June 2, 2020.
- Accepted September 14, 2020.
- Copyright © 2020 by The American Association of Immunologists, Inc.
References
In this issue
