Siglec-F Promotes IL-33–Induced Cytokine Release from Bone Marrow–Derived Eosinophils Independently of the ITIM and ITIM-like Motif Phosphorylation

Sialic acid–binding Ig-like lectin F (Siglec-F) is a surface receptor that belongs to the CD33 family of Ig-like lectins and is expressed on murine eosinophils, alveolar macrophages, and a few other cell types (1–3). Siglec-F and its human paralog Siglec-8 contain an ITIM and an ITIM-like motif in their cytoplasmic tail region (1, 2). Both receptors are able to induce apoptosis in eosinophils (4–6), while mechanistic differences were reported (7, 8). Studies with Siglec mutants revealed that the ITIMs play a critical role in the recruitment of the tyrosine phosphatases Src homology region 2 domain-containing phosphatase 1 (SHP-1) and SHP-2 that are able to inhibit activating signaling events (9–12). Siglec-F and Siglec-8 both bind to 6′-sulfo sialyl Lewis^X, which contains terminal sialic acids bound via α2,3 linkage to galactose-6-sulfate in glycoproteins (1). Although the mucins Muc5b and Muc4 were described to bind to Siglec-F (13), the identification and characterization of in vivo sialic acid ligands is still ongoing (13–15). It has been described that after ligand binding, Siglec-F limits eosinophil-mediated tissue damage during allergic inflammation by induction of apoptosis (4). However, this initial finding seems to be dependent on the experimental model. For example, Siglec-F–deficient mice showed enhanced lung eosinophilia in OVA-induced lung inflammation when the dissolved Ag was given intranasally, although this effect was not observed with aerosolized OVA, despite similar expression of Siglec-F ligands in both models (16). The mechanism of Siglec-F–induced apoptosis appears to be caspase dependent but relatively inefficient (8). Recent studies even describe Siglec-F–dependent enhanced effector molecule production. As such, Siglec-F knockdown on bone marrow–derived macrophages, which start to express Siglec-F after GM-CSF stimulation, leads to a reduced IL-4–induced phosphorylation of STAT6 and arginase-1 expression (17). In addition, combined stimulation of eosinophils with the alarmin IL-33 and anti–Siglec-F Ab caused a synergistically enhanced release of IL-4 and IL-13 compared with stimulation by either IL-33 or anti–Siglec-F treatment alone (18). This indicates that Siglec-F can act as an activating receptor besides its described proapoptotic function.

In this study, we analyzed the Siglec-F–mediated modulation of IL-33–induced chemokine and cytokine secretion from eosinophils. In vitro stimulation of bone marrow–derived eosinophils (BMDEs) with IL-33–induced Siglec-F upregulation and progressive secretion of IL-4, IL-13, MIP1-α/CCL3, and MIP1-β/CCL4 over several hours. IL-33 further impaired the proliferation of developing eosinophils in culture and inhibited the accumulation of eosinophils in the lung during Aspergillus fumigatus–induced allergic lung inflammation. Receptor cross-linking with anti–Siglec-F during IL-33 stimulation in vitro had little effect on viability, proliferation, and expression of activation markers but strongly enhanced cytokine/chemokine secretion. Interestingly, the IL-4/IL-13–activated transcription factor STAT6 was required for IL-33–induced secretion of IL-4, IL-13, CCL3, and CCL4, but anti–Siglec-F treatment could partially compensate for the STAT6 deficiency. Using retroviral reconstitution experiments, we demonstrate that the cytoplasmic tail, but not the tyrosine residues in the ITIM and ITIM-like motifs, is critical for the enhanced effector release on anti–Siglec-F cross-linking.

The following strains of mice were used on C57BL/6 background: IL-4eGFP reporter (4get_B6) (19), 4get_IL-5tg (20), IL-4/IL-13^−/− (21), ST2^−/− (22), MyD88^−/− (23), and CD45.1_B6-congenic mice. The following strains of mice were used on BALB/c background: 4get_BALB/c, 4get_STAT6^−/− (24), and CD45.1_BALB/c-congenic mice. To generate mixed bone marrow chimeras (MBMCs), we irradiated mice with 1100 rad, and 50:50 mixtures of either ST2^−/− and CD45.1-congenic or Siglec-F^−/− and CD45.1-congenic bone marrow cells were transferred by tail vein injection. MBMCs were kept with antibiotics in drinking water (2 g/l neomycin sulfate and 100 mg/l polymyxin B sulfate in drinking water; MilliporeSigma, Burlington, MA) for 8 wk after reconstitution and were analyzed after 8–10 wk. Mice were kept under specific pathogen-free conditions and maintained in the Franz-Penzoldt Center in Erlangen. All experiments were performed in accordance with German animal protection law and European Union guidelines 86/809 and were approved by the Federal Government of Lower Franconia.

BMDEs were generated as described before (25). In short, single-cell suspensions of bone marrow from femur and tibia were cultured in eosinophil culture medium (ECM; RPMI complemented with 20% (v/v) FBS, 25 mM HEPES, 55 μM 2-ME, 2 mM l-glutamine, 10 mM nonessential amino acids, 1 mM sodium pyruvate, 100 µg/ml streptomycin, 100 U/ml penicillin) at a density of 1 × 10⁶ cell/ml. For the first 4 days, 100 ng/ml SCF and 100 ng/ml FLT3-L (both PeproTech) were added to the culture and then changed to 10 ng/ml IL-5 (R&D Systems) for a further 10 days with media change every other day. Eosinophils were cultured in the presence of IL-5 to support their survival. At day 14, BMDEs were analyzed by flow cytometry (Siglec-F⁺, high side scatter, purity usually >90%). For ELISA and flow cytometry analysis, if not differently indicated, cells were seeded at a density of 0.2 × 10⁶ cells/ml and stimulated with 5 µg/ml purified rat anti-mouse Siglec-F Ab (clone E50-2440) or purified rat IgG2a isotype control (BD Pharmingen or Invitrogen) and/or 10 ng/ml recombinant murine IL-33 (R&D Systems) for 24 h. For Western blot, cells were seeded at a density of 1 × 10⁶ cells/ml and stimulated with 1 µg/ml anti–Siglec-F or isotype control Ab and 10 ng/ml IL-33.

Isolation of human eosinophils was performed on freshly isolated peripheral blood of healthy donors with informed consent and approval by the Ethics committee of the Faculty of Medicine at the Friedrich-Alexander Universität Erlangen-Nürnberg (no. 224_14B). About 50 ml whole blood was collected in an EDTA-containing tube, diluted 1:1 with PBS, and subjected to density gradient centrifugation with BioColl (1.077 g/ml; Bio&SELL) at 1000 × g for 30 min at room temperature. Supernatant was discarded, and the granulocyte-erythrocyte pellet was subjected twice to RBC lysis (15 and 10 min) in hypotonic lysis buffer (155 mM NH₄CL, 10 mM KHCO₃ in H₂O) on ice. Eosinophils were further separated from other granulocytes by magnet bead negative separation using the human Eosinophil isolation Kit (Miltenyi Biotec) according to the manufacturer’s instructions. Purity of the isolated eosinophils was determined directly after purification by flow cytometry–defining eosinophils as CCR3⁺Siglec-8⁺ singlets pregated on viable cells. Purity was at least >88% and on average 93% (Supplemental Fig. 2A). Stimulation experiments for activation marker flow cytometry and quantitative RT-PCR cells were performed in ECM supplemented with 10 ng/ml recombinant human IL-5 (R&D Systems). A total of 0.2 × 10⁶ cells/well in 200 µL were stimulated for 24 h with 5 µg/ml anti–Siglec-8 or isotype control Ab. A total of 10 ng/ml recombinant human IL-33 was optionally added.

Supernatants of IL-33–stimulated or unstimulated control BMDEs were collected over a kinetic from 30 min to 16 h and directly subjected to the ProcartaPlex Mouse Cytokine and Chemokine Panel 1A (36-plex) kit (eBioscience/Thermo Fisher Scientific). The assay was performed according to the manufacturer’s protocol and measured on a Bio-Rad Bio-Plex MAGPIX Multiplex Reader instrument.

Retroviral supernatants were generated in Phoenix or Plat-E cultures transfected with Siglec-F mutant sequences in a MSCV-IRES-Thy1.1 backbone and packaging helper plasmid pCLEco (Biocarta Europe). Siglec-F^−/− cells were transduced with retroviral supernatant in six-well plates at days 2 and 3 of BMDE generation. At day 13, Thy1.1⁺ cells were sorted. At day 14, 0.85 × 10⁶ cells were seeded in 100 µL ECM with 10 ng/ml IL-5 and stimulated with 5 µg/ml anti–Siglec-F (clone E50-2440), Rat IgG2a κ Isotype control (Becton Dickinson), and 10 ng/ml IL-33 (R&D Systems) as indicated in the figures for 24 h. Supernatants were used for ELISA.

A total of 2 × 10⁶ conidia of the ATCC 46645 strain of A. fumigatus were intranasally administered five times every 3–4 d as previously described (26). Mice were sacrificed on day 17 after initial application.

Either cultured cells or single-cell suspensions that were prepared using 70- or 100-μm cell strainers were analyzed. If necessary, erythrocytes were lysed with ACK buffer (0.15 M NH₄Cl, 1 mM KHO₃, 0.1 mM Na₂EDTA). If not indicated differently, stainings of up to 4 million cells were performed in 100 μl FACS buffer (PBS with 2% FCS and 0.1% NaN3) for 25 min at 4°C. Cells were pretreated with Fc receptor blocking Ab (anti-CD16/32, clone 2.4G2; BioXCell, West Lebanon, NH) and stained with fluorophore-coupled Abs. The following Abs for murine eosinophils were used: anti–Siglec-F BV421, Af647, PerCP-Cy5.5, PE or unlabeled clone: E50-2440, anti-B220 BV711 clone: RA3-6B2, anti–PD-L1 allophycocyanin clone: MIH5, anti-CD125 Af488 clone: T21, anti-Ly6G BUV395 clone: 1A8, anti-CD45.1 Af700 clone: A20, anti-CD11b BV785 clone: M1/70, anti-ST2 BUV421 clone: U29-93, and anti-CD11c BUV737 clone: N418 (Becton Dickinson); anti–Siglec-F unlabeled clone S17007L, anti-CD11c allophycocyanin-Cy7 clone: N418, anti-CD45.2 PE-Cy7 clone: 104, anti-CD115 PerCP-Cy5.5 clone: AF598, anti-CD193/CCR3 FITC or Af647 clone: J073E5, anti-CD11bs A700 clone: M1/70 (BioLegend); anti-Thy1.1 (anti-CD90.1) FITC or PerCP-Cy5.5 clone: HIS51, anti-CD62L PE-Cy7 clone: MEL-14, and anti-CD11b e780 clone: M1/70 (eBioscience); anti-CD101 PE clone: Moushi101, fixable viability dye in AmCyan (Invitrogen). For human eosinophils, the following Abs were used: anti-CD11b e780 clone: M1/70, anti-CD62L e450 clone: DREG-56, anti-CD193/CCR3 FITC clone: 5E8 (BioLegend); and anti–Siglec-8 PE clone: FAB7975P (BioLegend).

For proliferation assay, BMDEs of culture day 8 were seeded at 0.5 × 10⁶ cells/ml in 500 μl ECM medium in a 48-well plate and stimulated with 5 µg/ml rat IgG2a, к isotype control Ab or purified rat anti-mouse Siglec-F Ab (clone E50-2440) and optional 10 ng/ml recombinant murine IL-33 (R&D Systems) for 24 h. In addition, 10 μM EdU (Invitrogen) per well was added, and cells were incubated for a further 24 h at 37°C 5% CO2. Then cells were stained for surface Ags followed by EdU staining with the Click-iT Plus EdU Alexa Fluor 647 Kit (Invitrogen) according to the manufacturer’s instructions. For apoptosis assay, mature BMDEs were seeded at 0.2 × 10⁶ cells/ml in 200 μl BM medium in a 96-well plate and stimulated as in the proliferation assay described earlier for 24 h. For flow cytometry staining, Annexin binding buffer (10 mM HEPES, 140 mM NaCl, and 2.5 mM CaCl₂) was used instead of FACS buffer. Annexin V in FITC or allophycocyanin (ImmunoTools) was used, and DAPI (Sigma/Merck) was added to each sample directly before measuring in the flow cytometer.

For IL-4 ELISA, anti–IL-4 clones 11B11 (BioXCell) and biotinylated clone BVD6-24G2 (Southern Biotech, Birmingham, AL) were used as coating or detection Ab, respectively. For detection, phosphatase-coupled streptavidin and para-Nitrophenylphosphate substrate (Southern Biotech) were used. For IL-13 detection, an ELISA Development Kit (PeproTech) was used. For CCL3 and CCL4, ELISAs were also performed with commercial kits (CCL3/MIP-1 α DuoSet ELISA and Mouse CCL4/MIP-1 β DuoSet ELISA; R&D Systems).

Cells were lysed in radioimmunoprecipitation lysis buffer (1% Nonidet P-40, 50 mM Tris [pH 7.4], 0.15 M NaCl, 1 mM EDTA [pH 8.0], 0.25% deoxycholic acid) containing PhosSTOP and cOmplete Proteinase Inhibitor Cocktail (both Roche). A total of 5 µg protein was run on an SDS-PAGE and blotted to polyvinylidene difluoride (PVDF) membrane (Bio-Rad). Binding of primary Abs (all from Cell Signaling Technology) against phosho-STAT6 (Y641), STAT6 (either from Santa Cruz or Cell Signaling Technology), phospho-AKT (D9E), AKT (pan, C67E7), and β-actin (13E5) was detected by HRP-coupled donkey anti-rabbit (Rockland Immunochemicals) and the SignalFire Plus ECL reagent (Cell Signaling Technology) using the ChemiDoc imager (Bio-Rad). For incubation with a second primary Ab, membranes were stripped with the Restore PLUS Western blot Stripping Buffer (Thermo Fischer Scientific).

Freshly isolated human eosinophils were stimulated with 5 µg/ml anti-human Siglec-8 Ab (clone 7C9; BioLegend) or Purified Mouse IgG1 κ isotype control Ab (clone MOPC-21; BioLegend) and optionally 10 ng/ml recombinant human IL-33 (Life Technologies) in ECM, including 10 ng/ml recombinant human IL-5. After 4 h, cells were harvested in RLT buffer (Qiagen, Hilden, Germany) containing 0.1 M DTT and frozen overnight at −80°C. RNA was isolated with the RNeasy kit (Qiagen) according to the manufacturer’s protocol, and cDNA was synthesized with the SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). The SYBR Select Master Mix premix (Thermo Fisher Scientific) was used for quantitative PCRs, and cycle threshold values were measured on an Applied Biosystems ViiA 7 Real‐Time PCR System (Thermo Fisher Scientific) and normalized to GAPDH as a housekeeper gene. The following primer sequences were used: GAPDH-fw 5′- GGAGAAGGCTGGGGCTCATTTG-3′, GAPDH-rv 5′-GCATGGACTGTGGTCATGAGTCC-3′; CCL3-fw 5′-CCAGTTCTCTGCATCACTTGCTGC-3′, CCL3-rv 5′-CGCTTGGTTAGGAAGATGACACCG-3′; and CCL4-fw 5′-CGCTCTCAGCACCAATGGGC-3′, CCL4-rv 5′-GGAATACCACAGCTGGCTGGG-3′.

Eosinophils were derived from mice with systemic eosinophilia because of overexpression of IL-5 (4get_IL-5tg). Splenic single-cell suspension from male mice was treated with ACK buffer to lyse erythrocytes (0.15 M NH₄Cl, 1 mM KHO₃, 0.1 mM Na₂EDTA). Untouched eosinophils were sorted as SSC^high4get⁺ cells. Purity (Fixable Viability Dye⁻Siglec-F⁺CCR3⁺SSC^high) was >95%. Cells were seeded into 24-well plates (3 × 10⁶/ml) in 1 ml ECM with 10 ng/ml IL-5. Cells were stimulated with 5 µg/ml anti–Siglec-F (clone E50-2440) or isotype control (BD) and/or 10 ng/ml recombinant murine IL-33 (R&D Systems) for 4 h. RNA was isolated by phenol/chloroform extraction (83913; Sigma-Aldrich) according to the manufacturer’s instructions, except that 1/6 of recommended volumes were used, samples were sheared through 27G cannulas during initial lysis, and precipitation was performed at −80°C instead of −20°C. RNA was dissolved in diethyl pyrocarbonate H₂O. It was challenging to extract RNA from the RNase-rich eosinophils, especially after stimulation, and generated RNA had RNA Integrity Numbers between 5.7 and 7.8 (average 6.9; Agilent Bioanalyzer 2100). RNA was sent to Novogene Europe (Cambridge, UK) for sequencing as a service. There, library preparation was performed with NEBNext Ultra RNA Library Prep Kit for Illumina (NEB) following manufacturer’s recommendations, and paired-end sequencing was performed on an Illumina NovaSeq6000. Filtered clean reads were aligned to reference genome mm10 using STAR (v2.6.1d) (27). FeatureCounts (v1.5.0-p3) was used to generate counts used for downstream analysis in the laboratory (28). Data were loaded into R (3.5.3; The R Foundation for Statistical Computing, Vienna, Austria), and genes with zero counts were removed. Deseq2 (1.20.0) (29) was used to normalize counts and perform differential expression analysis without further shrinkage. Heatmaps of log2 normalized counts were drawn with the gplots package. For gene set enrichment, Bubble GUM software (1.3.19) (30) based on gene set enrichment analysis (GSEA) (31) with normalized count values as input and recommended settings (default values but permutation type set to “gene_set”) was used. For the analysis, HALLMARK gene sets (32) of MSigDB database version 6.2 were combined with custom added gene sets of IL-33– or IL-4–stimulated eosinophils (GSE43660; significantly altered and log2 fold change > 0.25) (33). GSEA built-in conversion of mouse to human gene identifiers was used if necessary (31). Data are available via the GEO database (accession no. GSE166968; https://www.ncbi.nlm.nih.gov/geo/).

Siglec-F cDNA was derived from splenic cells of an IL-5tg mouse (identical with NCBI CCDS ID 21167.1). The primer was 5′-G GCGGCCGC GCCACC ATGCGGTGGGCATG-3′ (SigF-for; features in order separated by space: guanine, NotI, Kozak, complementary sequence); 5′-G GTCGAC TCAGCACTTGTGGATCTTGATCTCTGTG-3′ (SigF-rev; guanine; SalI, complementary sequence). Overlap PCR with the cloned Siglec-F as a template generated a Siglec-F version with mutated ITIM and ITIM-like motif. The primer for overlap PCR 1 (mutated triplets highlighted by extra spaces) was 5′-GAGCCTGAACTCCAC TTT GCCTCCCTCTCCTTCC-3′ (SigF-for-ITIMmut), 5′-TCAGCACTTGTGGATCTTGATCTCTGT GAA TACAG 3′ (SigF-rev-ITIM-like-mut). In parallel, another PCR with the SigF-for primer and a complementary reverse version of the SigF-for-ITIMmut primer was performed to generate the second template. Both templates were used in an overlap PCR with the SigF-for and the SigF-rev primer to finalize the Siglec-F sequence with mutated ITIM and ITIM-like motifs. To generate the Siglec-F version with a truncated intracellular tail, we used SigF-for primer combined with 5′ G GTCGAC TCA CTTCACTGTGAAAAAGATG-3′ (guanine, SalI, stop codon, complementary sequence). The sequential tail truncation variants were generated analogously to the outlined tail truncation variant using corresponding reverse primers. Siglec-F variants were cloned into an MSCV-IRES-Thy1.1 via NotI and SalI restriction sites.

Statistical analysis was performed with GraphPad PRISM 5 and Sigmaplot 12.3 except for RNA sequencing analysis. Statistical tests were used as indicated in the figure legends.

The alarmin IL-33 is known to activate eosinophils, although the spectrum of secreted cytokines and chemokines has not been analyzed in detail. To address this point, we analyzed supernatants of IL-33–stimulated BMDEs (culture exemplary visualized in Supplemental Fig. 1A) by a Luminex bead array. The secretion of the proinflammatory cytokine IL-6 and the IL-1 family member IL-18, as well as the type 2 immunity-associated cytokines IL-4 and IL-13, was strongly induced on IL-33 stimulation (Fig. 1A). The chemokines MCP-1/CCL2, MIP-1α/CCL3, MIP-1β/CCL4, and MIP-2/CCL8 that contribute to the recruitment of immune cells were also elevated (Fig. 1A). Kinetic ELISA experiments further revealed that secretion of IL-4, IL-13, CCL3, and CCL4 was already detectable at 2 h after stimulation, and the concentrations significantly increased after 8 and 16 h compared with unstimulated controls (Fig. 1B). Simultaneously, IL-33 induced the upregulation of Siglec-F on the cell surface (Fig. 1C). Because Siglec-F has been described as an inhibitory and proapoptotic receptor (4, 5), these results raise the question how the potentially opposing processes of enhanced activation/cytokine production by IL-33 and inhibitory effects from Siglec-F engagement are balanced and integrated to mediate an appropriate immune response.

Next, we investigated the combined effect of IL-33 and Siglec-F stimulation on eosinophil survival. BMDEs were stimulated with IL-33 or anti–Siglec-F alone, or with a combination of both. After 24 h, the viability (percent of DAPI⁻Annexin⁻ cells, gating exemplary visualized in Supplemental Fig. 1B) was only mildly affected by anti–Siglec-F treatment (Fig. 2A, 2B). This is in accordance with published in vitro data that used the same anti–Siglec-F Ab clone (4, 7, 8). Related to this, we determined whether Siglec-F mediates inhibition of eosinophil proliferation. We measured proliferation by the uptake of EdU (gating exemplary visualized in Supplemental Fig. 1B) starting at day eight of the eosinophil culture, when cells still proliferate. Cells were stimulated for 48 h, and EdU was added for the last 24 h of stimulation. Although IL-33 treatment caused profound inhibition of BMDE proliferation, there was only a mild inhibitory effect caused by additional anti–Siglec-F stimulation (Fig. 2C, 2D).

To further analyze the regulation of eosinophilia by IL-33 and Siglec-F in vivo, we generated MBMCs and subjected them to intranasal A. fumigatus infection to elicit allergic inflammation in the lung. First, we analyzed MBMCs generated with 50% CD45.2_ST2^−/− (IL-33R deficient) and 50% congenic CD45.1_wild-type donors. Although the ratio of eosinophils in the lung remained ∼50:50 in naive MBMCs, the ratio changed to a higher percentage of ST2^−/− cells after A. fumigatus infection (70:30) (Fig. 3A). Representative flow cytometry plots for naive and A. fumigatus–infected mice highlight A. fumigatus–induced eosinophil (SiglecF⁺Ly6G⁻ gate) recruitment to the lung (Fig. 3B). Further subgating of eosinophils by congenic markers reveals the fraction of eosinophils derived from ST2^−/− or wild-type bone marrow (Fig. 3B).

Next, we analyzed eosinophils in MBMCs generated with 50% CD45.2_Siglec-F^−/− and 50% CD45.1_wild-type donors. We used a gating strategy to identify eosinophils as CD11b⁺SSC^hi cells without using Siglec-F as an actual marker (Supplemental Fig. 1C). We observed an expected increase of lung eosinophils from 8 ± 5% SD in naive to 46 ± 14% SD in A. fumigatus–infected chimeras (data not shown). Although we expected a 50:50 ratio of bone marrow–derived cells after reconstitution, we observed a ∼20:80 (Siglec-F^−/−/wild-type) ratio of eosinophils, neutrophils, and B cells in the lung of naive and A. fumigatus–infected MBMCs (Fig. 3C, 3D). This indicates that Siglec-F may play a general role for bone marrow reconstitution but does not significantly reduce the number of eosinophils in the inflamed lung, which does not fit with a proapoptotic function. The subset of CD101⁺ inflammatory eosinophils (34) was slightly increased within the Siglec-F^−/− fraction in uninfected MBMCs, but this difference disappeared after A. fumigatus infection (Fig. 3E, 3F).

In conclusion, IL-33 inhibits eosinophil proliferation in vitro, and IL-33–responsive eosinophils have a competitive disadvantage to establish lung eosinophilia after A. fumigatus infection, which we have previously shown to promote eosinophilopoiesis in the bone marrow (35). In contrast, anti–Siglec-F caused only minor effects on eosinophil survival and proliferation in vitro. In accordance, Siglec-F–deficient eosinophils did not outcompete wild-type eosinophils in A. fumigatus–induced lung eosinophilia of MBMCs. Therefore, the results do not support a strong proapoptotic role of Siglec-F under these conditions.

Because we observed increased Siglec-F expression after IL-33 stimulation and in inflamed lungs that was accompanied by enhanced cytokine/chemokine secretion on IL-33 stimulation, we further investigated how anti–Siglec-F regulates the IL-33–induced cytokine/chemokine secretion and expression of activation markers on the cell surface. Therefore, BMDEs were stimulated with IL-33, anti–Siglec-F, or a combination of both. Concentrations of the type 2 immunity–associated effector molecules IL-4, IL-13, CCL3, and CCL4 were then measured in supernatants by ELISA. Double stimulation of eosinophils with IL-33 and anti–Siglec-F Ab (either clone E50-2440 or clone S17007L) had a strong synergistic effect compared with stimulation with IL-33 or anti–Siglec-F alone (Fig. 4A, 4B, and data not shown). In contrast, anti–Siglec-F treatment did not significantly alter the IL-33–induced upregulation of CD11b, CD11c, CD101, or ST2 but slightly enhanced the downregulation of CD62L (Fig. 4C). To further investigate whether the IL-33–induced change of activation marker expression is a cell-intrinsic effect, we analyzed IL-33–stimulated BMDE cultures generated with 50% CD45.2_ST2^−/− and 50% congenic CD45.1_wild-type cells derived from MBMC mice. The activation markers CD11b and CD101 were indeed expressed at lower levels on ST2^−/− eosinophils as compared with wild-type eosinophils after IL-33 stimulation, whereas expression of CD62L that is downregulated after IL-33 stimulation and CCR3 remained higher on ST2^−/− eosinophils (Fig. 5). This result proves that cell-intrinsic ST2-mediated signaling events are directly required for IL-33–induced changes of the activation marker profile, and the effect cannot be caused in a paracrine manner by soluble factors released from bystander wild-type eosinophils. In addition, we found that activation marker expression and mediator secretion after IL-33 stimulation require the ST2-associated signaling adaptor molecule MyD88 (Supplemental Fig. 3A–C). In summary, anti–Siglec-F treatment promotes IL-33–elicited and MyD88-dependent mediator secretion, but anti–Siglec-F Ab by itself does not affect expression of activation markers.

To determine whether Siglec-F–mediated enhanced effector release of IL-33–stimulated eosinophils is transferrable to the human system, we targeted Siglec-8, a described functional paralog of Siglec-F in humans, on eosinophils isolated from the peripheral blood (Supplemental Fig. 2A). We confirmed published Siglec-8–induced cell death by treatment with a commonly used anti–Siglec-8 Ab (clone 7C9) (Supplemental Fig. 2B) (5) but also upregulation of the activation marker CD11b (Supplemental Fig. 2C) (7, 36). As a further sign of activation, we observed CD62L downregulation on human eosinophils, while CCR3 expression was not altered (Supplemental Fig. 2C). CCL3 and CCL4 gene expression were also upregulated after anti–Siglec-8 Ab stimulation (Supplemental Fig. 2D). However, combined anti–Siglec-8 Ab and IL-33 treatment at indicated time points did not induce a synergistic upregulation of activation markers or chemokines beyond single stimulations (Supplemental Fig. 2). Furthermore, IL-4 or IL-13 was not detected on combined anti–Siglec-8 Ab and IL-33 treatment, or treatment with one or the other alone (data not shown). These findings indicate that Siglec-8 does not promote IL-33–elicited eosinophil activation but, in contrast with Siglec-F, can activate eosinophils directly as reported before (36).

Next, we extended our investigation of the synergistic IL-33 and anti–Siglec-F effect on potentially involved signaling pathways. We analyzed phosphorylation of the IL-4/IL-13–activated transcription factor STAT6 and the serine/threonine kinase AKT that is involved in regulation of cell metabolism and survival. As expected, AKT phosphorylation was induced by IL-33, but anti–Siglec-F treatment did not modulate this response (Fig. 6A). However, anti–Siglec-F treatment enhanced and prolonged the IL-33–induced phosphorylation of STAT6 (Fig. 6A). To further investigate whether IL-4/IL-13–mediated activation of STAT6 was indeed required for mediator secretion or expression of activation markers, we used BMDEs from STAT6^−/− mice. The induction of activation markers in the absence of STAT6 was still functional with subtle STAT6-dependent alterations, and only moderate (for CD11b, CD11c) or minor (for ST2, CD101, CD62L, CCR3) alterations were observed compared with wild-type (Supplemental Fig. 3D). However, the IL-33–induced secretion of IL-4, IL-13, CCL3, and CCL4 was almost completely abrogated in STAT6^−/− BMDEs (Fig. 6B). Only low levels of IL-13 were detected in supernatants of IL-33–stimulated STAT6^−/− eosinophils. This demonstrates that IL-33 signaling does not directly result in pronounced secretion of IL-4, IL-13, CCL3, and CCL4, which suggests that IL-33–elicited autocrine IL-4/IL-13 signaling through STAT6 is required for substantial secretion of all four mediators. Consistently, CCL3 and CCL4 concentrations were also reduced in IL-33–stimulated IL-4/IL-13^−/− BMDE cultures (Fig. 6C), whereas only minor changes of activation marker expression could be observed (Supplemental Fig. 3E).

Interestingly, although anti–Siglec-F alone does not induce effector release, anti–Siglec-F and IL-33 cotreatment was able to elicit low-level secretion of all four mediators in STAT6^−/− BMDE cultures, indicating that the STAT6 dependency for IL-33–induced effector secretion could be partially restored by Siglec-F signaling (Fig. 6B). This finding led us to further investigate Siglec-F–mediated transcriptional changes to understand which pathways might promote enhanced effector release from IL-33–stimulated eosinophils.

In addition to signaling analysis on the protein level, we analyzed how IL-33 and anti–Siglec-F double stimulation impacts transcription by RNA sequencing. Although anti–Siglec-F alone barely changed transcription (only Etv5, Peg10, Spred1, Spred2, Tnf, Tram1, and Trem14 were weak but significantly upregulated), IL-33 induced the expression of 224 genes >2-fold compared with the isotype control. These included the cytokines/chemokines Il6, Il13, Ccl3, Ccl4, and Cxcl2. Combined IL-33 and anti–Siglec-F treatment led to an even more pronounced upregulation of the highly induced genes. In addition, a higher total number of genes was induced >2-fold compared with IL-33 treatment alone (454), but changes were moderate for most of them (Fig. 7A). To determine differentially expressed genes specific for the IL-33 and anti–Siglec-F double treatment, we determined genes that were significantly different versus all of the single/control stimulations (isotype, isotype + IL-33, or anti–Siglec-F) (Fig. 7B). Among the upregulated genes were hypoxia-associated genes (Fam162a, Egln2, Ankrd37, Hilpda, Ftl1, Prdx1) and likely hypoxia-induced metabolism/glycolysis-associated genes (Pfkl, Tpi1, Pgk1, Pgm2), as well as a gene encoding for the hypoxia-induced, proinflammatory cytokine migration inhibitory factor (MIF). How exactly Siglec-F signaling confers a hypoxia-associated gene expression profile is beyond the scope of this article and remains to be determined in future studies. We further noticed an enhanced expression of Tarm1, which encodes for a costimulator of cytokine secretion (37). This could at least partially explain the anti–Siglec-F–mediated increase of cytokine/chemokine concentrations in supernatants of IL-33–stimulated eosinophils. Other upregulated genes on anti–Siglec-F and IL-33 costimulation included proapoptotic and antiapoptotic genes (Bcl2a1b, Bcl2a1d, Mien1, Bnip3), integrin-related/adhesion-related Cd53, the gene for the surface receptor CD52 involved in cell activation and proliferation, genes encoding for transcription factors and transcription factor binding proteins (Eef1e1, Batf3, Scand1), the protease inhibitor coding gene Cstb, the G protein signaling-related gene Rgs10, and ribosomal proteins (Nhp2, Rps27l, Rps14) (Fig. 7C). The downregulated genes included repressors of NF-κB (Mturn) and TGF-β (Ldlrad4) pathways. To analyze the data beyond the top regulated genes, we performed a GSEA on HALLMARK gene sets (32), where we also compared IL-33 and anti–Siglec-F treatment with the other conditions. Myc target genes and gene sets associated with Mtorc signaling, oxidative phosphorylation, generation of reactive oxygen species, glycolysis, and adipogenesis were all significantly upregulated after double stimulation compared with any of the single treatments (Fig. 7D). As expected, a huge proportion of gene sets was enriched on stimulation with IL-33 alone. Notably, most sets enriched by IL-33 showed at least a tendency to further enrichment by additional anti–Siglec-F treatment (i.e., the HYPOXIA gene set).

In line with HALLMARK analysis results, effects on published IL-33 target genes (33) were also elevated when eosinophils were treated with anti–Siglec-F + IL-33 compared with isotype + IL-33 treatment. In contrast, IL-4 signature genes were enriched when stimulation contained IL-33, but additional significant effects on combined anti–Siglec-F treatment were not observed (Fig. 7D). Notably, IL-4 expression itself was also not significantly altered by any of the stimulations applied in the experiment. Therefore, increased IL-4 secretion observed on anti–Siglec-F + IL-33 treatment did not result in a stronger IL-4–mediated transcriptional response compared with IL-33 alone. In contrast, IL-13 transcription is upregulated by IL-33 stimulation, but there is no significant further upregulation on additional anti–Siglec-F treatment.

In summary, anti–Siglec-F treatment generally does very little by itself but elevate gene expression induced by IL-33, but together with IL-33 stimulation also induces expression of additional, i.e., hypoxia/glycolysis-associated, genes.

The anti–Siglec-F–mediated enhanced effector release from IL-33–stimulated eosinophils suggested a critical role for the cytoplasmic tail of Siglec-F in this process. The tail of Siglec-F contains one ITIM and one ITIM-like motif that mediate downstream signaling. To analyze whether these motifs or other parts of the tail are required for enhanced effector release in the context of IL-33 and anti–Siglec-F double stimulation, we generated BMDEs from Siglec-F^−/− bone marrow cells that were retrovirally complemented with wild-type or mutant Siglec-F. Either both the ITIM and ITIM-like motif were mutated by exchange of their tyrosine residues to phenylalanine (ITIMmut) or the cytoplasmic tail was completely deleted (stop codon behind K463; Taildel) (Fig. 8A). The transduced Siglec-F^−/− BMDEs successfully expressed all constructs on the cell surface, demonstrating that the tail is not required for surface expression of Siglec-F (Fig. 8B). Reconstitution with wild-type Siglec-F did not spontaneously modulate IL-33–induced secretion of IL-4, IL-13, CCL3, and CCL4 (Fig. 8C, Supplemental Fig. 4A). However, enhanced secretion occurred after additional anti–Siglec-F treatment of wild-type and ITIMmut samples (with the exception of CCL4), but not with Taildel samples (Fig. 8C–E, Supplemental Fig. 4A, 4B). In an attempt to further specify which region of the cytoplasmic tail enhances effector release, we generated two additional truncation mutants. The first lacks the terminal part of the tail that includes the ITIM-like motif and the ITIM motif (stop codon behind D528; ITIMdel), and the second one retains only a short part of the cytoplasmic tail (stop codon behind S493; Shorttail). In both mutants, the synergistic effect of IL-33 and anti–Siglec-F treatment was lost (Supplemental Fig. 4C). This indicates that the last 42 amino acids deleted in the ITIMdel mutant are required to mediate the synergistic effect. In conclusion, the terminal part of the Siglec-F tail is critical for the elevated production of IL-4, IL-13, and CCL3 in activated eosinophils, but tyrosines in its ITIM and ITIM-like motifs are dispensable. Our study shows an important activating function of the Siglec-F tail independently of known cytoplasmic signaling motifs generally associated with inhibitory functions. Therefore, Siglec-F should not simply be regarded as a proapoptotic receptor but also as an activating modulator of cytokine and chemokine secretion from eosinophils.

Eosinophils are involved in allergic responses as tissue-damaging effector cells. Therefore, various treatments have been developed that target eosinophils to ameliorate diseases such as allergy and asthma. These include eosinophil depletion or the activation of inhibitory receptors such as Siglec-8 on human and as a model Siglec-F on murine eosinophils (4–6, 8, 38–44). Based on their ITIM and ITIM-like motifs, both receptors are thought to mediate inhibition and apoptosis of eosinophils (4, 5). However, at least in the murine system, it was shown that the induction of apoptosis by Siglec-F was not always effective (4, 7, 8) and was strongly dependent on the chosen disease model (16). Such context-dependent differences are currently not well understood. A recent study with bone marrow–derived macrophages, which are able to express Siglec-F after GM-CSF stimulation, described enhanced Siglec-F–dependent STAT6 phosphorylation and in the presence of IL-4 also arginase 1 induction (17). In contrast with the described functions of Siglec-F, this might argue for promotion of type 2 immunity by Siglec-F, rather than functional inhibition of the macrophages. Along the same line, our group found that IL-33 and Siglec-F costimulation lead to elevated IL-4 and IL-13 secretion by eosinophils in vitro (18). We expand these initial findings now to increased secretion of CCL3 and CCL4 chemokines. Both chemokines can also be sensed by murine eosinophils (45) and mediate effector cell recruitment to sites of inflammation, where they can contribute to tissue damage. We further describe that the apoptosis induction by Siglec-F is only a moderate effect as also seen by others (4, 7, 8) and largely independent of the IL-33 costimulation. Therefore, the elevated effector release after IL-33 and Siglec-F costimulation cannot simply be explained by passive release caused by Siglec-F–induced apoptosis. Notably, IL-33 inhibited proliferation measured at days 7–8 of our BMDE cultures, while others reported an IL-33–dependent expansion of bone marrow eosinophil precursors and in turn mature eosinophils (46). However, BMDE cells probably develop uniformly in culture and early precursor stages that likely proliferate and drive an IL-33–mediated eosinophil expansion that might not be present in the middle of the BMDE culture. Therefore, IL-33 might modulate eosinophil proliferation dependent on the eosinophil developmental stage. We also find no obvious proapoptotic function of Siglec-F in vivo by analyzing A. fumigatus–induced lung eosinophilia in competitive MBMCs. In combination, our findings in the context of the current literature point to a more diversified role of Siglec-F with inhibitory and activating aspects that are regulated in a context-dependent way. It might be that Siglec-F elevates the immune response until a signal, maybe via a feedback mechanism, induces inhibition/apoptosis to prevent overwhelming immune responses as suggested in the literature (4). How these findings relate to the human system remains incompletely understood because cell death induction by Siglec-8 is described to be more pronounced and mechanistically different from the induction by Siglec-F (8). This might indicate a higher inhibitory capacity of Siglec-8, despite the described analogy of Siglec-8 and Siglec-F. It has even been described that cytokines such as IL-5 and IL-33 that enhance effector secretion in combination with Siglec-F in the murine system in contrast enhance Siglec-8–dependent cell death induction of human eosinophils (36, 47, 48). Because induction of cell death by Siglec-8 includes activation of signaling pathways and reactive oxygen species effector production, Siglec-8 was already reported as an activating receptor in this regard (36). We also observed activation-associated surface receptor regulation but no enhancement of MIP1-α, MIP1-β, IL-4, or IL-13 chemokine/cytokine release on anti–Siglec-8 Ab treatment of IL-33–stimulated human eosinophils that would point to synergistic activating functions of IL-33 and Siglec-8 in the chosen setup (47).

The effector release after IL-33 or combined IL-33 and anti–Siglec-F treatment is almost exclusively dependent on MyD88 as a main mediator of IL-33 signaling (23, 49, 50), which we show in MyD88^−/− BMDE cultures. Downstream we find IL-33 to induce STAT6 phosphorylation. This has been observed before and was linked to an IL-33–mediated IL-4 release that induces STAT6 phosphorylation via an autocrine IL-4 loop (33). Compatible with this model, the further enhanced IL-4 release observed on combined IL-33 and anti–Siglec-F treatment probably also induces the elevated and prolonged STAT6 phosphorylation. Using BMDEs from STAT6-deficient mice, we confirm that STAT6 signaling enhances effector release. However, a Siglec-F–induced effector release is also observed in STAT6-deficient BMDEs, albeit at a lower level, leaving the possibility of additional STAT6-independent pathways of Siglec-F–induced secretion of effectors.

The transcriptome analysis revealed hypoxia- and glycolysis-related genes, such as the cytokine MIF, to be highly upregulated by Siglec-F costimulation. MIF promotes inflammation, eosinophil maturation, chemotaxis, and activation in type 2 immunity-related diseases (51), and it may contribute to enhanced IL-13 secretion as suggested by others for murine asthma models (52, 53). Tarm1, another gene upregulated by Siglec-F costimulation, has been shown to enhance inflammatory cytokine secretion by macrophages and neutrophils on LPS stimulation (37), suggesting that it may also promote Siglec-F–mediated enhanced effector release in eosinophils. On a broad perspective, a higher number of genes is differentially expressed on Siglec-F costimulation, and GSEA additionally suggests that Siglec-F primarily elevates enrichment of gene sets induced by IL-33. Therefore, Siglec-F is able to function as a costimulatory molecule for IL-33–induced expression changes.

Our mutation analysis shows that tyrosine phosphorylation of the inhibitory ITIM and ITIM-like motifs of the cytoplasmic tail are not mandatory for enhanced effector release. We mutated their tyrosine residues to phenylalanine, which has been described as effective to also prevent phosphorylation-independent SHP-1 and SHP-2 binding for another Siglec (12) and to mediate loss of inhibitory capacity of Siglec-8 on mast cells (54). Therefore, we likely suppress most of the described signaling capacity of the tail in the ITIM mutant. Nevertheless, enhanced effector release of IL-33 and anti–Siglec-F treatment was absent only when the whole tail of Siglec-F was deleted also excluding that the remaining domains act as coreceptors for other proteins to mediate effector release. However, enhanced effector release was still seen in the ITIM mutant. This again suggests ITIM-independent signaling capacity that induces activating functions of Siglec-F. However, we cannot exclude structural changes of the tail deletion mutant that might lead to an altered localization or interaction behavior.

In summary, Siglec-F is able to enhance IL-33–dependent secretion of the cytokines IL-4 and IL-13 and the chemokines CCL3 and CCL4 mediated by its cytoplasmic tail but independent of tyrosine phosphorylation in the ITIM and ITIM-like motifs, likely related to increased expression of hypoxia-induced proinflammatory molecules and enhanced/prolonged STAT6 phosphorylation.

We thank Roland Lang for providing MyD88^−/− bone marrow; Sven Krappmann, Sebastian Schrüfer, and Michaela Dümig for providing A. fumigatus conidia; Christian Schwartz for providing human IL-33; Mark Gresnigt for advice on human eosinophil isolation; and the sequencing core unit at the Institute of Human Genetics of the University Hospital Erlangen, especially Arif Ekici, for help to determine RNA quality.

B.S.B. receives remuneration for serving on the scientific advisory board of Allakos, Inc. and owns stock in Allakos. B.S.B. is a coinventor on existing Siglec-8–related patents and thus may be entitled to a share of royalties received by Johns Hopkins University during development and potential sales of such products. B.S.B is also a cofounder of Allakos, Inc., which makes him subject to certain restrictions under university policy. The terms of this arrangement are being managed by Johns Hopkins University and Northwestern University in accordance with their conflict of interest policies. The other authors have no financial conflicts of interest.