Siglec-15 on Osteoclasts Is Crucial for Bone Erosion in Serum-Transfer Arthritis

Rheumatoid arthritis (RA) is a chronic inflammatory autoimmune disease that primarily affects the joints. With a prevalence of 0.5–1%, RA is a very frequent autoimmune disease worldwide (1). However, causes and interactions of various cell types and mechanisms during RA have not been fully identified. Although successful treatment strategies for RA were developed [e.g., TNF inhibitors (2)], they are associated with several side effects and therapy failures in numerous patients, and therefore, new targets for therapy need to be found. At early stages of RA patients suffer from joint pains, swelling, and stiffness, among other symptoms. In the further course of disease, these symptoms are followed by a synovitis accompanied by cartilage and bone destruction (3, 4). In late stages of RA, patients with a chronic course of disease often also develop systemic osteoporosis (5). The exact mechanism of pathologic bone loss during RA is still not fully clarified. Initially, cell types, such as macrophages, monocytes, and T cells infiltrate the synovial tissue (4, 6, 7). IL-6, TNF, and other proinflammatory cytokines are secreted by these cell types, which, inter alia, amplify osteoclast differentiation and activation caused by upregulation of the receptor activator of NF-κB ligand (RANKL) and M-CSF (6–8). These two cytokines, M-CSF and RANKL, are key regulators of osteoclastogenesis, the maturation process from mononuclear into multinuclear bone resorbing osteoclasts (9).

Osteoclasts are considered to be primarily responsible for bone loss during RA. For one thing, multinucleated cells expressing markers, such as tartrate-resistant acid phosphatase (TRAP) and cathepsin K, which are located at the sites of bone lesions in RA patients, have been identified as osteoclasts (10, 11). Furthermore, increased concentrations of RANKL, M-CSF, or TNF-α, which are cytokines supporting differentiation of precursor cells into osteoclasts, have been found in inflamed tissue of RA patients (12–15). In this context, blocking of RANKL, for instance, significantly reduces bone erosions and systemic bone loss in RA patients (16). Studies using different mouse models of arthritis support the primary role of osteoclasts in pathologic bone loss, as mice lacking osteoclasts develop no bone erosions during arthritis (17, 18). Additionally, expression of DAP12 and FcRγ are upregulated in the active RA tissue compared with healthy tissue (19). These two adaptor proteins are also known to be important for osteoclastogenesis (20). However, because of their short extracellular domains, these adaptor proteins are thought to interact with other immunoreceptors on the plasma membrane of osteoclasts.

Sialic acid–binding Ig-like lectin 15 (Siglec-15) has been identified as such a transmembrane receptor, which can interact with DAP12 and also has a regulatory role in osteoclastogenesis (21–23). Siglec-15 belongs to the family of conserved Siglecs and is found in mouse and human (21). It is expressed on the surface of osteoclasts, and the expression is upregulated during osteoclastogenesis in response to RANK signaling and the subsequent induction of transcription factor NFAT2 activity (22–24). A lack of Siglec-15 results in defective osteoclast maturation in vitro and, moreover, in an impaired bone resorbing function that underlines the important regulatory role for Siglec-15 in osteoclastogenesis and bone remodeling (22, 23, 25). Both interaction with the adaptor protein DAP12 and Siglec-15 ligand binding play a crucial role in promoting maturation (23, 26). In concert with the adaptor protein DAP12, Siglec-15 regulates osteoclast development via RANKL-induced phosphorylation of PI3K, p85, Akt, and Erk (26, 27). In vivo studies using Siglec-15^−/− mice show impaired osteclastogenesis, resulting in an increased trabecular bone mass, especially in the secondary spongiosa (25, 26). However, in the primary spongiosa, there seems to be a compensatory mechanism via the OSCAR/Fcγ pathway (26). This might be the reason that Siglec-15 deficiency results only in a mild osteopetrotic phenotype. Similarly, treatment of healthy wild-type mice or rats with an anti–Siglec-15 Ab results in increased bone mass and increased mineral density (27, 28).

It is known that the mechanisms of pathologic bone loss differ from healthy bone remodeling, but whether Siglec-15 is also involved in the pathologic bone loss has to be investigated. There are two publications concerning the role of Siglec-15 in osteoporosis. In 2015, a role for Siglec-15 in postmenopausal osteoporosis was demonstrated, as Siglec-15^−/− female mice were protected from bone loss after ovariectomy (29). In a second study, an anti–Siglec-15 Ab therapy was successfully tested in growing rats and could therefore represent a potential treatment of juvenile osteoporosis (28). However, in the context of RA, the role of Siglec-15 has not been extensively studied. So far, only one study was performed using Siglec-15^−/− mice in an Ag-induced arthritis (AIA) model. In this AIA model, a similar inflammation and comparable bone erosion within the knee joints of Siglec-15^−/− and wild-type mice was observed. The authors concluded that Siglec-15 is not involved in joint inflammation or destruction in AIA (30). Instead, they found differences in periarticular osteoporosis between Siglec-15^−/− and wild-type mice.

Apart from the role of Siglec-15 in arthritis, its expression pattern on cell types other than osteoclasts also remains largely unknown. Initial analyses regarding the expression pattern of Siglec-15 showed coexpression with CD68 in histological sections of human spleen and lymph nodes (21), indicating Siglec-15 expression on macrophages. Additionally, studies presented its expression on tumor-associated macrophages in various human tumor tissues in immunohistology (31). In vitro experiments in this study suggested a role for Siglec-15 in enhancing TGF-β secretion from macrophages in the tumor environment, which may contribute to tumor progression (31). More recently, immunosuppressive characteristics of Siglec-15 on macrophages in vivo were revealed, and it was therefore postulated as a target for cancer immunotherapy (32). The immune-suppressive function of Siglec-15 suggests that it may bind to ligands on tumor cells. Siglecs bind to sialic acids in specific linkages on glycoproteins. With its extracellular ligand binding domain, human Siglec-15 was reported to bind preferentially the sialyl-Tn (Neu5Acα2-6GalNAcα) structure, whereas murine Siglec-15 recognizes both the sialyl-Tn structure as well as α2–3–linked 3′sialyllactose (Neu5Acα2-3Galβ1-4Glc) (21). The sialyl-Tn moiety is a carbohydrate structure, which is restricted to tumor cells (24). A recent study did not confirm the preferential binding of human Siglec-15 to sialyl-Tn but demonstrated a broader recognition of sialylated glycans by human Siglec-15 (33).

In this study, we used newly generated Siglec-15^−/− mice to study the role of Siglec-15 in arthritis. We demonstrated an important role for Siglec-15 in bone erosion during serum-transfer–induced arthritis. Furthermore, the expression pattern of Siglec-15 on murine and human immune cells was examined in detail with a new mAb.

A targeting vector was generated containing two loxP sites flanking the exons 2–5 of the Siglec-15 gene, a neomycin- and a HSV-thymidine-kinase cassette. JM8A3 embryonic stem cells (C57BL/6 background with corrected Agouti gene) (34) were transfected with the targeting vector DNA, selected with G418 as well as ganciclovir, and screened by PCR and Southern Blot. Positive embryonic stem cells, in which correct homologous recombination into the Siglec-15 gene locus could be confirmed, were injected into blastocysts isolated from pregnant C57BL/6 mice. The blastocysts were implanted into pseudopregnant female mice, which gave birth to chimeric mice. The male mice with the highest percentage of chimerism were crossed with B6 wild-type female mice, and the offspring were further screened by PCR for the targeting sequence as an indicator for successful germline transmission. To delete the neomycin cassette, mice were bred with FLP-expressing mice and afterward crossed with EIIA-cre mice to generate Siglec-15 deficiency. These Siglec-15–deficient mice have a pure C57BL/6 background. All animal experiments were approved by the local authorities (Regierung von Unterfranken).

For the generation of a monoclonal anti-murine Siglec-15 Ab, a Siglec-15–deficient mouse was immunized i.p. with 100 μg of murine Siglec-15 Fc protein precipitated in alum (Alu-S-Gel; Serva). The Fc protein consists of extracellular Ig1–2 domain of the murine Siglec-15 fused to the Fc part of a human IgG1 H chain and was produced by LEC-1 CHO cells (the Fc protein construct was kindly provided by T. Angata, Academia Sinica, Taipei, Taiwan). Six weeks after immunization the mouse was boosted with the same amount of Siglec-15 Fc protein–Alum complex. Successful anti–Siglec-15 Ab production was confirmed by testing serum, which was collected 7, 14, and 21 d after the second immunization on a Siglec-15–transfected DT40 cell line. Five days prior the hybridoma fusion, the mouse was challenged again with 15 μg of Fc protein in sterile PBS i.v. For the hybridoma fusion, the spleen was isolated from the immunized mouse, and a cell suspension was prepared in serum-free medium (RPMI 1640; PAN-Biotech) and centrifuged (1400 rpm, 10 min, 4°C). Additionally, a comparable amount of myeloma SP2/0 cells was pelletized via centrifugation (1400 rpm, 10 min, 4°C). The cell pellets were pooled in 10 ml of serum-free medium. After centrifugation (1000 rpm, 10 min, 37°C) the supernatant was discarded, and 2 ml of prewarmed polyethylene glycol (37°C; Roche) was added slowly within 60 s to induce the fusion of the two cell types. A 90-s incubation at 37°C was performed before the reaction was stopped by adding 25 ml of serum-free medium within a time interval of 6 min. The pelletized cells (1000 rpm, 37°C, 10 min) were resolved, seeded on ten 96-well plates, and selected in medium containing 10% FCS (PAN-Biotech), 2 mM l-Glutamine (PAN-Biotech), 100 U/ml penicillin/streptomycin (PAN-Biotech), 50 μM 2-ME (PAN-Biotech), 1% IL-6 supernatant, and 1× hypoxanthine-aminopterin-thymidine (Sigma-Aldrich) for 10 d. Supernatants of the cell clones were tested for anti–Siglec-15–positive hybridomas by staining of Siglec-15–transfected DT40 cells and detection with Fc-specific anti-mouse IgG Ab (PE labeled; The Jackson Laboratory). One anti–Siglec-15 Ab–producing hybridoma clone could be identified (MK4.5), which was expanded. For Ab production, the MK4.5 cells were cultured in IgG-free Hybridoma Medium 6 (Bio & SELL) plus 1% supernatant from an IL-6–producing cell line for 13 d, and afterward, the Ab was purified from the supernatant via protein G–Sepharose column and labeled with biotin or FITC.

Murine osteoclasts were generated as described (35). Erythrocyte lysis was done for 2 min at room temperature using a lysis buffer consisting of 115 mM NH₄Cl, 10 mM NaHCO₃, and 1 mM EDTA in H₂O. Bone marrow cells were plated in a six-well cell culture dish (Greiner Bio-One) in medium (4 mM l-glutamine, 100 U/ml penicillin/streptomycin, and MEM-α; Life Technologies) containing 1% FCS and incubated at 37°C and 5% CO₂ for 14–16 h. Afterward, the nonadherent cells were plated in a concentration of 1 × 10⁶ cells/ml in osteoclast medium containing 10% FCS (Life Technologies) as well as 30 ng/ml murine M-CSF (PeproTech) and 10 ng/ml mRANKL (PeproTech). The size of the chosen well plate depended on the experiment the osteoclasts were used for (2 × 10⁵ for 96-well plate, which were used for TRAP staining; 1 × 10⁶ for 24-well plate was used for immunofluorescence and flow cytometry analyses). Medium was changed after 3 d, and 10–24 h after medium change, the osteoclasts were stained with an Acid Phosphatase Leukocyte Kit (catalog no. 387A; Sigma-Aldrich) for TRAP staining or were stained for flow cytometry or immunofluorescence.

Osteoclasts from PBMCs were generated as previously described (35). For the gradient to isolate PBMCs, Lymphoprep (STEMCELL Technologies) was used. The size of the chosen well plate depended on the experiment for which the osteoclasts were made (3 × 10⁵ for 96-well plate, which were used for TRAP staining; 1.5 × 10⁶ for 24-well plate was used for immunofluorescence and flow cytometry analyses). Medium was changed on day 3 and 6. On day 7, the osteoclasts were stained with the Acid Phosphatase Leukocyte Kit (catalog no. 387A; Sigma-Aldrich) for TRAP or were stained for flow cytometry or immunofluorescence.

PBMCs were isolated from the buffy coat the same way as for osteoclast generation (35). For the gradient to isolate PBMCs, Lymphoprep (STEMCELL Technologies ) was used, and all washing steps were done with 5 mM EDTA/PBS. A total of 1.5 × 10⁸ PBMCs were stained (30 min, 4°C, gently mixing several times) in 4.5 ml of sorting buffer (5 mM EDTA [Life Technologies]/PBS [Life Technologies] plus 2% heat-inactivated FCS [PAN-Biotech]) added with anti-CD14-PECy7 Ab (HCD14; BioLegend). After two washing steps with 5 ml sorting buffer, the cells were resuspended in sorting buffer in a concentration of 5 × 10⁷ cells/ml, and CD14-high cells were sorted in macrophage medium (MEM-α, [Life Technologies], 4 mM l-glutamine, [Life Technologies], 100 U/ml penicillin/streptomycin [Life Technologies], 10% heat-inactivated FCS [PAN-Biotech]) using a FACSAria (BD Biosciences). After centrifugation (1300 rpm, 4°C, 5 min), cells were resuspended in medium added with 100 ng/ml human M-CSF (PeproTech) and seeded on 24-well plates (Greiner Bio-One) for 7 d. On day 3 and 6, the medium was changed.

For immunofluorescence analysis, the murine bone marrow cells or human PBMCs were seeded on 11-mm glass slides in 24-well plates and, afterward, differentiated into osteoclasts as described above. On day 5 (murine) or day 7 (human), adherent cells were washed with 1× PBS (Life Technologies) for 10 min, followed by a 10 min fixation step with 70% methanol (precooled) on −20°C. The cells on glass slides were air-dried before they were washed twice with PBS for 5 min. This was followed by an incubation step with 0.01% Tween 20 (Carl Roth) in PBS for 10 min and, again, two washing steps for 5 min. Cells on glass slides were treated with 50 mM NH₄Cl for 5 min, washed two times, then were blocked for 1 h at room temperature (the blocking solution contained 0.1% OVA [Applichem], 1.1% fish gelatin [Sigma-Aldrich] in 0.15 M PBS [pH 7.4]). To prevent unspecific binding, the murine cells on glass slides were incubated with Fc receptor–blocking Abs against FcγRIV, FcγRII, and FcγRIII (clone 9E9 and 2.4G2) for 15 min in blocking solution. The human cells were incubated with human Fc receptor blocking solution (Human TruStain FcX; BioLegend). The incubation with unlabeled anti-murine Siglec-15 (MK4.5 clone; murine, 5.5 μg/ml; human, 7 μg/ml) in blocking solution occurred over night at 4°C. After three washing steps (5 min), the cells were stained with DAPI (0.4 μg/ml; Sigma-Aldrich) and goat anti-mouse IgG-AF568 (1:500; Thermo Fisher Scientific) in blocking solution for 2 h at room temperature. After three washing steps in PBS (5 min) and one last wash in double distilled H₂O, the cells on glass slides were embedded using 5 μl of Aqua-Poly/Mount (Polysciences). Samples were examined with an Axio Imager M2 equipped with an ApoTome.2 module by using ZEN blue 2012 software (ZEISS, Oberkochen, Germany). For comparison, images of Siglec-15^+/+ and Siglec-15^−/−osteoclasts were taken with the same exposure time as stacks of multiple optical sections. The same procedure was used for human-derived osteoclasts stained with anti–Siglec-15 (MK4.5) or corresponding isotype control. Projections were calculated with ZEN blue 2012 software (ZEISS), adjusted equally for contrast and brightness using Adobe Photoshop CS6 (Adobe Systems, San Jose, CA) and arranged using Corel Draw X9 (Corel, Munich, Germany).

In the serum-transfer mouse model of arthritis, serum from transgenic K/BxN mice (36) was pooled. Siglec-15^+/+ and Siglec-15^−/− mice were injected i.p. with 300–400 μl/male or 200–300 μl/female. The severity of arthritis was determined via a scoring system from 0 to 3 per paw following 10 d (no swelling = 0, mild swelling = 1, moderate swelling = 2, severe swelling of paw and ankle = 3). On day 10, the mice were sacrificed and bones from hind legs were isolated and fixed in 4% paraformaldehyde (PFA; Roth) in PBS at 4°C. Depending on the experimental setup, the bones were then used for computed tomography (CT) analyses or decalcified for histology analyses.

Magnetic resonance imaging (MRI) was performed on an ultra-high-field scanner (ClinScan 7 Tesla; Bruker, Ettlingen, Germany) using an axial Turbo spin echo T2-weighted sequence with the following specifications: slice thickness, 500 μm; echo time, 89; repetition time, 8000; and voxel size, 59 × 59 × 50 μm. For MRI analysis, an OsiriX DICOM Viewer (Aycan, Rochester, New York) was used in combination with Chimaera’s segmentation tool (Chimaera, Erlangen, Germany). In that way, the fraction of T2 hyperintense tissue was determined in the paw (volume of T2 hyperintensity in mm³ divided by volume of the paw in mm³) to semiquantitatively assess tissue edema in the respective hind leg.

Hind limbs were isolated and fixed in 10% formalin solution (Sigma-Aldrich) or alternatively in 4% PFA (Roth) in PBS at 4°C. CT was performed on a Siemens Inveon CT scanner (Siemens Healthineers, Erlangen, Germany) with the following settings: tube voltage, 80 kV; tube current, 500 μA; exposure time, 1200 ms; binning, 2 × 2; charge-coupled device size, 1152 × 2048 pixels; field-of-view, 29.19 × 51.89 mm; effective pixel size, 50.67 μm; and scan time, 22 min. For CT image analysis, OsiriX DICOM Viewer (Aycan) was used in combination with Chimaera’s segmentation tool (Chimaera). In that way, Hounsfield unit (HU) values of the metaphyseal femur were measured to investigate bone density by placing a region of interest in the metaphysis (in vivo steady-state analysis). For the ex vivo serum-transfer–induced arthritis analysis, volumes of the tarsal and metatarsal osteolyses in the hind paws were determined by defining a threshold of HU values between 1000 and 2200 HU. The resulting volume of bone lesions was determined in mm³.

After fixation with 4% PFA/PBS the bones were decalcified in 14% (w/v) EDTA (Sigma-Aldrich) buffer (pH 7.2, adjusted by NH₄OH addition) for 14 d and embedded in paraffin. For histology analyses of the paws, removal of the nails was important. Five micrometer paraffin sections were stained with either Safranin O for evaluation of cartilage destruction or H&E for identification of inflammatory cell infiltration. To measure bone erosion areas and the number of osteoclasts, TRAP staining was used. For evaluation, the OsteoMeasure analysis system (OsteoMetrics) was used.

To analyze different murine cell populations of naive mice by flow cytometry, single-cell suspensions of spleen, bone marrow, or the peritoneal cavity were prepared in PBS (Life Technologies) using a 70-μm cell strainer (Greiner Bio-One). In analyses of liver cells, a perfusion with PBS was performed beforehand. Depletion of erythrocytes was achieved by incubation of the cells with ACK buffer (115 mM NH₄Cl, 10 mM NaHCO₃, and 1 mM EDTA in H₂O) for 2 min at room temperature. Cell concentration and absolute numbers of living cells were determined using trypan blue staining of the cells and counting viable cells via Neubauer cell counting chamber. Afterward, the cells were washed with FACS buffer (0.1% BSA, 2 mM EDTA, and 2 mM sodium azide in PBS) and stained with Abs conjugated to biotin, FITC, PE, PerCP-Cy5.5, allophycocyanin, Alexa Fluor 488, allophycocyanin/Cy7, PE-Vio770, and Alexa Fluor 700. Additionally, to avoid unspecific binding via the Fc part of the Abs, an Fc block was added (2.4G2, hybridoma purified in our laboratory) and, depending on the experimental setup, DAPI (0.2 μg/ml; Sigma-Aldrich) or fixable viability dye (Invitrogen) was used to exclude dead cells during measurement. For staining of the different cell populations, the following Ab clones were used: anti-CD4 (GK1.5), anti-CD8a (53-6.7), anti-TCRβ (H57-597), anti-B220 (RA3-6B2), anti-IgM (II/41), anti-IgD (11-26c.2a), anti-CD11c (N418), anti–Ly-6G (1A8), anti-CD11b (M1/70), anti–Siglec-F (E50-2440), anti–Gr-1 (RB6-8C5), anti-CD62L (MEL-14), anti-F4/80 (BM8), anti-Tim4 (F31-5G3), and anti-CD45.2 (104). Biotinylated Abs were detected by staining with streptavidin conjugated to one of the fluorochromes mentioned above. After staining, the cells were washed with FACS buffer and resuspended in either FACS buffer or FACS buffer containing 2% PFA. Data were acquired via CytoFLEX S (Beckman Coulter) or FACSCalibur (Becton Dickinson) and analyzed using FlowJo software.

To analyze human cell populations by flow cytometry, depletion of erythrocytes was achieved by incubation of human blood with 9× vol H₂O for 30 s at room temperature. Subsequently PBS (Life Technologies) was added, and after a centrifugation (5 min, 1300 rpm, 4°C), cells were resuspended in 1× PBS. Cell concentration was determined using trypan blue staining of the cells and counting viable cells via Neubauer cell counting chamber. Depending on the experimental setup, 3 × 10⁵–1 × 10⁶ human peripheral blood leukocytes were used per staining. Cells were washed with FACS buffer (0.1% BSA, 2 mM EDTA, and 2 mM sodium azide in PBS) and were incubated with Fc receptor–blocking solution (Human TruStain FcX; BioLegend) in FACS buffer according to the manufacturer’s instructions. Abs used for staining of human cells were conjugated to biotin, FITC, PE, PE-Cy5, PE/Cy-7, allophycocyanin, allophycocyanin–Fire 750, allophycocyanin/Cy-7, BV421, BV605, BV510, PE/Dazzle, or Alexa Fluor 700. For staining of Siglec-15, biotinylated MK4.5 clone was used. The same was true for the isotype control (MG1-45; BioLegend), which was biotinylated in our laboratory the same way as anti–Siglec-15 Ab (MK4.5). For staining of the different cell populations, the following Ab clones were used: anti-CD3 (UCHT1), anti-CD19 (HIB19), anti-CD11c (3.9), anti-CD33 (HIM3-4), anti-CD45 (HI30 or 2D1), anti-CD16 (3G8), anti-CD14 (M5E2 or HCD14), anti-CD66c (B6.2CD66c), anti-CD56 (MEM-188), and HLA-DR (L243; BioLegend). Biotinylated Abs were detected by staining with streptavidin conjugated to allophycocyanin. After staining, the cells were washed with FACS buffer and resuspended in FACS buffer. Data were acquired via CytoFLEX S (Beckman Coulter) and analyzed using FlowJo software.

For flow cytometry analyses of in vitro–generated osteoclasts, the adherent osteoclasts were removed from 24-well plates on day 5 via incubation with 0.5 mM EDTA/PBS for 15 min and pipetting up and down. To avoid clumps and clogging of the cytometer, the cells were flushed through a 100-μm cell strainer (Greiner Bio-One). Approximately 1 × 10⁶ cells were stained at 4°C in 25 μl of FACS buffer (0.1% BSA plus 2 mM EDTA plus 2 mM sodium azide in PBS) containing corresponding staining Abs as well as murine Fc receptor–blocking Abs (clone 2.4G2, hybridoma purified in our laboratory and clone 9E9, purified from hybridoma cells). Abs that were used are as follows: anti–Siglec-15 Ab (clone MK4.5; our own hybridoma), anti-F4/80 Ab (BM8), anti-OSCAR Ab (11.1CN5), and DAPI. To remove unbound Abs, the cells were washed with FACS buffer, and after centrifugation (1300 rpm, 5 min, 4°C), the cells were resuspended in FACS buffer for measurement. Data were acquired via CytoFLEX S (Beckman Coulter) and analyzed using FlowJo software.

On day 7, in vitro–generated macrophages were removed from 24-well plates via incubation with 5 mM EDTA for 15 min and pipetting up and down. To avoid clumps and clogging of the cytometer, the cells were flushed through a 100-μm cell strainer (Greiner Bio-One). Afterward, the cells were incubated with Fc receptor–blocking solution (Human TruStain FcX; BioLegend) in FACS buffer (0.1% BSA plus 2 mM EDTA plus 2 mM sodium azide in PBS) according to the manufacturer’s instructions. After one washing step using FACS buffer, cells were stained with either FACS buffer containing anti-CD14-PE/Cy7 (HCD14; BioLegend) (fluorescence minus one), anti-CD14-PE/Cy7 plus IgG1 isotype control (MG1-45; BioLegend)-biotin, or anti-CD14-PE/Cy7 plus anti–Siglec-15 (MK4.5; our own hybridoma)–biotin for 30 min at 4°C. Cells were washed with FACS buffer and incubated with FACS buffer containing streptavidin–allophycocyanin (PharMingen) for 30 min, 4°C. The cells were washed twice and were resuspended in FACS buffer for flow cytometry measurement (CytoFLEX S; Beckman Coulter), and data were analyzed using FlowJo software.

Statistical analyses were performed using GraphPad Prism software. Depending on distribution of values, for single comparison, an unpaired t test or Mann–Whitney U test was used to evaluate significance. Statistical data are presented as mean ± SD or median ± SD, as indicated in the legends below each figure.

To investigate the physiological role of Siglec-15 in mouse arthritis models, we generated Siglec-15^−/− mice (Supplemental Fig. 1). In contrast to other Siglec-15–deficient mouse models, our new mouse model allows a conditional knockout of Siglec-15, depending on the Cre deleter strain (Supplemental Fig. 1A). In this study, only complete Siglec-15 knockout mice generated by a universal Cre deleter, on a pure C57BL/6 background were used and characterized. Because this is a newly generated Siglec-15 knockout mouse (designated Siglec-15^−/−), we first analyzed the role of Siglec-15 in osteoclastogenesis and bone remodeling, as Siglec-15 is important for osteoclast maturation (22, 23). For this purpose we isolated bone marrow cells from our Siglec-15^−/− mice and differentiated them in vitro into osteoclasts. TRAP staining of these in vitro–generated osteoclasts (Fig. 1A) demonstrates that Siglec-15^−/− cells fail to develop into multinucleated mature osteoclasts. In contrast to osteoclasts generated from Siglec-15^+/+ cells, we only detected very few and small TRAP-positive cells with a small number of nuclei (Fig. 1A). Quantification by counting mature osteoclasts confirmed significantly reduced multinucleated osteoclasts in cells lacking Siglec-15 (Fig. 1B).

To clarify whether this impaired osteoclastogenesis leads to a reduced bone resorption in vivo and therefore results in an altered bone remodeling, the bone density of naive Siglec-15^−/− mice was analyzed via CT. Siglec-15^−/− mice show a significant increase in bone density compared with wild-type mice within the metaphyseal region of the femur (Fig. 1C, 1D). These findings are in accordance with previous results of other Siglec-15^−/− mouse strains (25, 26).

Our main purpose of the generation of the Siglec-15^−/− mice was to study its role in arthritis. Thus far, it has not been conclusively determined whether Siglec-15 is also involved in the pathologic bone loss during arthritis. We used the serum-transfer model of arthritis, which is known for severe inflammation and bone erosion. For this purpose Siglec-15^−/− mice and control mice were challenged with the K/BxN serum (36, 37). First of all, we followed the course of disease over 10 d but did not observe differences in the clinical arthritis score in Siglec-15^−/− mice compared with wild-type mice (Fig. 2A). On day 10, the mice were sacrificed, and metatarsal regions of hind legs were analyzed by histological stainings to examine the inflammation area, cartilage destruction (loss of proteoglycan), and bone destruction, as well as the corresponding number of osteoclasts. The results show no alterations regarding inflammation area (determined by H&E) or loss of proteoglycan (determined by Safranin O staining) in Siglec-15^−/− mice when compared with wild-type mice (Fig. 2B–D). TRAP staining revealed massive bone erosion in Siglec-15^+/+ mice at day 10 but hardly any bone loss in Siglec-15^−/− mice (Fig. 2B, 2E). Furthermore, quantification of osteoclasts at the sites of bone erosion revealed a significantly reduced number of osteoclasts in Siglec-15^−/− mice compared with control mice (Fig. 2F). These data demonstrate that the lack of Siglec-15 protects mice from bone loss during serum-transfer–induced inflammatory arthritis. Furthermore, we analyzed the mice on day 10 after K/BxN serum transfer by CT and MRI. CT analyses of the metatarsal regions of hind legs showed more osteolysis in Siglec-15^+/+ mice and less in Siglec-15^−/− mice (Fig. 3A, 3B). Similar to the histology data, the MRI analyses on day 10 showed no differences in paw swelling in Siglec-15^−/− compared with Siglec-15^+/+ mice (Fig. 3C).

Because the surface expression pattern of Siglec-15 on immune cells other than osteoclasts has not been fully clarified, we generated a monoclonal anti–Siglec-15 Ab (mAb) by immunizing Siglec-15^−/− mice with a murine Siglec-15 Fc protein. One hybridoma clone (MK4.5) was identified by staining the Siglec-15–transfected cell line DT40 and was then tested on in vitro–generated osteoclasts. Although very large multinucleated osteoclasts (>100 μm) could not be analyzed by flow cytometry because of their size, smaller aggregates of mature osteoclasts could be easily handled by the cytometer. We developed a method for flow cytometry staining of in vitro–generated osteoclasts by using the markers F4/80, the forward scatter (FSC), combined with Siglec-15. A defined population of F4/80^low, Siglec-15⁺ mature osteoclasts was present in Siglec-15 wild-type cells but absent in cells lacking Siglec-15 (Fig. 4A). Fig. 4B also clearly shows that anti–Siglec-15 mAb MK4.5 stains Siglec-15 on FSC^high F4/80^low cells, which are mature osteoclasts, but does not stain precursor cells (FSC^low F4/80^neg) or immature osteoclasts (FSC^high F4/80^high) (Fig. 4B). Staining of in vitro–generated human osteoclasts, differentiated from human PBMCs, with anti–Siglec-15 mAb MK4.5 revealed a cross-reaction of our Ab clone with human Siglec-15 (Fig. 4D). For this reason, our newly generated anti–Siglec-15 Ab MK4.5 can be used to detect both murine and human Siglec-15. Because of their size, large mature multinucleated osteoclasts are difficult to analyze by flow cytometry. Therefore, our anti–Siglec-15 mAb MK4.5 was additionally used to stain mature osteoclasts in immunofluorescence (Fig. 4C, 4E). The immunofluorescence pictures present convincing staining of Siglec-15 on both murine and human mature osteoclasts, whereas Siglec-15^−/− cells or corresponding isotype control showed no signal. This demonstrates that our anti–Siglec-15 mAb MK4.5 is a reliable new Ab to stain osteoclasts from mice and humans and works both in flow cytometry and immunofluorescence.

Siglec-15 expression has also been previously described on macrophage subpopulations and dendritic cells (DCs) (21). To determine the Siglec-15 expression pattern on immune cells, we used our anti–Siglec-15 Ab MK4.5 and analyzed various murine cell populations in different organs regarding Siglec-15 expression. For this purpose, murine lymphocytes, DCs, granulocytes, and monocytes, as well as macrophages isolated from blood, bone marrow, thymus, lymph node, spleen and peritoneal cavity, were tested for surface Siglec-15 expression via flow cytometry (Fig. 5). As depicted in Fig. 5, none of these cell types expresses Siglec-15 in any analyzed organ, with the exception of large peritoneal macrophages, which show a slight but reproducible Siglec-15 expression (Fig. 5E) but a lot less than mature osteoclasts (compare with Fig. 4). However, macrophages in other organs, such as spleen or liver, did not show Siglec-15 expression on their surface in naive Siglec-15 wild-type mice compared with Siglec-15^−/− mice (Fig. 5E). In addition to that, there was also no Siglec-15 expression detectable on murine bone marrow–derived macrophages when using the MK4.5 Ab clone in flow cytometry (Fig. 5F). These results show that Siglec-15 is exclusively expressed on the surface of osteoclasts and very weakly expressed on CD11b^high, F4/80^high peritoneal macrophages.

The consequence of the Siglec-15 deficiency on immune cell populations has not been described in detail so far. Therefore, we analyzed B cell subpopulations (Supplemental Fig. 2A), T cells (Supplemental Fig. 2B), DCs (Supplemental Fig. 2C), macrophages and neutrophils (Supplemental Fig. 2D, 2E), and classical as well as nonclassical monocytes (Supplemental Fig. 2F) of Siglec-15–deficient and control mice. Cell numbers of all of these immune cells were not altered in any of the isolated organs of Siglec-15^−/− mice. This indicates that Siglec-15 deficiency does not affect cell populations other than osteoclasts, which is compatible with the expression pattern of Siglec-15.

Human Siglec-15 has recently been described as a potential new target for cancer immunotherapy, as it is upregulated on tumor-infiltrating macrophages (32). Therefore, a thorough surface expression analysis of Siglec-15 on human immune cells is important and has not been done so far. Thus, we used our mAb MK4.5 to analyze Siglec-15 surface expression on various PBL populations. To obtain a reliable negative control for staining, a mouse IgG1 isotype control was biotinylated with the same protocol as our anti–Siglec-15 mAb MK4.5. We determined Siglec-15 expression on human lymphocytes and myeloid cell types, gated as shown in Supplemental Fig. 3. T cells, B cells (Fig. 6A), and NK cells (Fig. 6B) did not express any Siglec-15. However, Siglec-15 is expressed on human neutrophils (Fig. 6C, 6G), and furthermore, apart from one donor, classical and nonclassical monocytes had mostly high Siglec-15 expression (Fig. 6D, 6G). Expression on DCs was low (Fig. 6E). We also generated macrophages from CD14⁺-sorted PBMCs by differentiation with M-CSF for 7 d. These in vitro–generated macrophages showed high expression levels of Siglec-15 (Fig. 6F). Our results demonstrate that Siglec-15 is more broadly expressed on different human myeloid cell types of peripheral blood compared with the mouse, in which it is almost exclusively found on osteoclasts.

It has previously been shown that Siglec-15 is an important regulator of osteoclastogenesis and healthy bone remodeling (22, 23, 25, 26). However, it has not yet been clarified whether Siglec-15 is also involved in the pathologic bone erosion process. To address this, we generated Siglec-15–deficient mice and analyzed them in the K/BxN serum-transfer–induced arthritis model. We showed for the first time, to our knowledge, that there is a clear role of Siglec-15 on osteoclast-mediated pathological bone erosion in inflamed joints in an arthritis mouse model. Interestingly, the arthritis score of Siglec-15^−/− mice was comparable to wild-type mice after serum transfer. However, histological analyses revealed a significant reduction in bone erosion areas, as well as a reduced number of osteoclasts within these areas of the inflamed joints in our Siglec-15^−/− mice, whereas the inflammation area and loss of proteoglycan was comparable to wild-type mice. These findings show that the inflammation and bone erosion in arthritis are not necessarily tightly linked and suggest a specific role of Siglec-15 in the latter process. Our interpretation of these findings is that the inflammatory process within the joints triggers local osteoclast differentiation that causes pathological local bone erosion and this osteoclast differentiation is blocked in Siglec-15^−/− mice; therefore, these mice are protected from the local bone loss. Osteoclast maturation involves fusion of osteoclast precursor cells to multinucleated mature osteoclasts. Siglec-15 binding to sialic acid–containing ligands on neighbor cells during this fusion process may be involved in this, but the mechanism is still unclear.

Another group did not observe altered osteolysis directly in arthritic knee joints of Siglec-15^−/− mice compared with wild-type mice in the AIA model. They reported a comparable inflammation and bone erosion between wild-type and Siglec-15^−/− mice (30). Instead, they found differences in periarticular osteoporosis in bone areas around the inflamed joint. This is likely a secondary effect but not the primary erosion typically seen in RA patients. We also worked with the AIA model in Siglec-15^−/− mice and observed a low degree of inflammation and a relatively weak knee-swelling during AIA that are not comparable to the strong swelling of paws during serum-transfer arthritis model. Furthermore, we observed more severe bone erosion areas in the serum-transfer model than in AIA. Therefore, we think a certain threshold of inflammation and subsequent osteoclast maturation is necessary to observe the bone-eroding function of Siglec-15 during arthritis. Of course, the two arthritis models differ in their setup. In contrast to AIA, which is an actively induced model in which mice are preimmunized with methylated BSA and the effector phase is induced by direct methylated BSA injection into the knee, the serum transfer is a passive arthritis model. In this model, serum from K/BxN mice containing high levels of anti-glucose 6 phosphate isomerase Abs is injected into recipient mice (37, 38). This leads to a spontaneous inflammatory arthritis in the paws, which is mainly mediated by the alternatively activated complement and Fc-receptors (39). Furthermore, innate immune cells, such as neutrophils, are recruited to the affected paws, driving a strong inflammatory reaction (40). Siglec-15 might be directly or indirectly involved in these pathways induced by the K/BxN serum transfer.

Additionally, we were able to confirm the defective osteoclastogenesis in vitro and the mild osteopetrotic phenotype in vivo, which have been demonstrated in two other Siglec-15–deficient mouse models (25, 26). This confirmation is important, as these phenotypes have now been found in three independently generated Siglec-15–deficient mouse lines. It has previously been postulated that Siglec-15 might be a new target for osteoporosis. Indeed, Siglec-15–deficient mice were resistant to ovariectomy-induced osteoporosis (29). In healthy young mice and rats in vivo, anti–Siglec-15 Ab administration led to increased bone mass due to inhibited osteoclast differentiation (27, 28). This suggests that Siglec-15 represents a promising target for the therapy of postmenopausal or juvenile osteoporosis (28, 29). An anti–Siglec-15 Ab treatment also seems to be promising for arthritis models and will be tested by us in the future.

By immunizing our Siglec-15^−/− mice with a Siglec-15 Fc protein, we generated an anti–Siglec-15 mAb named MK4.5. This anti–Siglec-15 Ab was useful for investigating the expression pattern of Siglec-15 on various cell types. With this Ab, we demonstrate a high expression level of Siglec-15 on in vitro–generated murine- and human-derived mature osteoclasts of mice and humans. Siglec-15 expression on other cell types has been less clear. Because an early study about Siglec-15 with a polyclonal Siglec-15 Ab revealed coexpression of Siglec-15 with DC-SIGN and CD68 on human histological section of spleen and lymph nodes, it has been speculated that Siglec-15 might be expressed on macrophages and/or DCs (21). However, in this study, via flow cytometry, we were able to exclude surface expression of murine Siglec-15 on myeloid cell types, such as DCs and macrophages in the spleen and liver, and on bone marrow–derived macrophages. Interestingly, no Siglec-15 expression was detectable on murine monocytes, although they are the precursor cells of osteoclast maturation during inflammatory arthritis (41). Moreover, only a very low Siglec-15 expression was detectable on large peritoneal macrophages by flow cytometry. Recently, by use of another Siglec-15 mAb, a similar, very low or absent Siglec-15 expression on murine macrophages in different tissues was reported (32). Furthermore, we could now also demonstrate a broad expression of Siglec-15 on human myeloid cells. This is important as approaches to target Siglec-15 on tumor-infiltrating macrophages are discussed. Also, targeting Siglec-15 on osteoclasts as a new osteoporosis treatment in humans seems to be feasible. Therefore, it is important to know which other immune cell types would be affected by such a treatment. We also found high Siglec-15 levels on human monocyte-derived macrophages, which is similar to data published before (32).

Because of the much more restricted expression pattern of murine Siglec-15, we expected no effects on cell populations other than osteoclasts in our Siglec-15^−/− mice. In accordance, our data showed no alterations in total immune cell numbers in Siglec-15^−/− mice when compared with wild-type mice.

In conclusion, we have clarified the expression pattern of Siglec-15 on osteoclasts and several other immune cell populations of mouse and human with a newly generated mAb. Importantly, to our knowledge, we are the first to show Siglec-15 plays a role in pathologic bone erosion in a mouse arthritis model by controlling osteoclast maturation. This directly suggests Siglec-15 as a suitable target for therapy of pathologic bone loss during arthritis. Such therapy could be done by anti–Siglec-15 Abs, as it has been demonstrated for osteoporosis before. Although our new Ab did not block osteoclast differentiation, we are in the process of generating new mAbs that could serve such a function in the future.

We thank Takashi Angata for providing an expression construct for the Siglec-15 Fc-protein. Furthermore, we thank Claudia Koller, Uwe Appelt, Markus Mroz, Heike Danzer, and Jutta Jordan for technical assistance. We also thank Fabian Andes, Ulrike Steffen, Anja Lux, Sibel Kara, and Andreas Gießl for support with technical questions.

The authors have no financial conflicts of interest.