Soluble CD93 lectin-like domain sequesters HMGB1 to ameliorate inflammatory diseases

Rationale: CD93, a C-type lectin-like transmembrane glycoprotein, can be shed in a soluble form (sCD93) upon inflammatory stimuli. sCD93 effectively enhances apoptotic cell clearance and has been proposed as an inflammatory disease biomarker. The function of sCD93 involved directly in inflammation remains to be determined. Herein, we attempted to examine the hypothesis that sCD93 might sequester proinflammatory high-mobility group box 1 protein (HMGB1), exerting anti-inflammatory properties. Methods: Different forms of soluble recombinant human CD93 (rCD93) were prepared by a mammalian protein expression system. rCD93-HMGB1 interaction was assessed using co-immunoprecipitation and solid-phase binding assays. Effects of soluble rCD93 were evaluated in HMGB1-induced macrophage and vascular smooth muscle cells (VSMC) activation and receptor activator of nuclear factor-κB ligand (RANKL)-induced osteoclastogenesis, CaCl2-induced and angiotensin II-infused abdominal aortic aneurysm (AAA) formation and ovariectomized-induced osteoporosis in mice. Results: Protein binding studies revealed that soluble rCD93, via the lectin-like domain (D1), can bind to HMGB1 and intercept HMGB1-receptor interaction. Soluble rCD93 containing D1 inhibited HMGB1-induced proinflammatory cytokine production and intracellular mitogen-activated protein kinase (MAPK)/nuclear factor (NF)-κB activation in macrophages and VSMCs, thereby attenuating CaCl2-induced and angiotensin II-infused AAA models. During osteoclastogenesis, RANKL stimulated HMGB1 secretion that promoted RANKL-induced osteoclastogenesis in return. Soluble rCD93 containing D1 impeded RANKL-induced osteoclastogenic marker gene expression and intracellular MAPK/NF-κB signaling, thereby mitigating ovariectomized-induced osteoporosis. Conclusion: These findings demonstrate the therapeutic potential of soluble recombinant CD93 containing D1 in inflammatory diseases. Our study highlights a novel anti-inflammatory mechanism, i.e., sequestration of HMGB1, through which sCD93 prevents HMGB1-receptor interaction on effector cells and alleviates inflammation.

In the studies presented here, we attempted to investigate the hypothesis that sCD93 might interact with HMGB1, a crucial proinflammatory mediator, thereby preventing it from engaging its receptors (e.g., RAGE) and hampering downstream inflammatory signaling. For this purpose, soluble recombinant human CD93 (rCD93) ectodomain proteins were prepared. We examined whether soluble rCD93 binds to HMGB1 and suppresses HMGB1-induced inflammatory events. In addition, in vivo studies were performed to evaluate whether soluble rCD93 mitigates HMGB1-driven inflammatory diseases, including experimental CaCl2-induced and angiotensin II (AngII)-infused abdominal aortic aneurysm (AAA) and ovariectomized (OVX)-induced osteoporosis in mice. These findings may help gain insight into the functional significance of sCD93 in inflammatory disorders.
The enzyme-linked immunosorbent assay (ELISA) solid-phase binding assay was performed to assess the direct binding of soluble rCD93 and HMGB1. In an assay with mole-equivalent rCD93D123, rCD93D1, and rCD93D23 immobilized on the plate, the binding of HMGB1 to rCD93D1 and rCD93D123 was increased by HMGB1 in a concentration-dependent manner, whereas the binding of HMGB1 and rCD93D23 was persistently low ( Figure 1C). Consistently, in an assay with incremental concentrations of rCD93 proteins immobilized on the plate, the binding of HMGB1 to rCD93D1 and rCD93D123, but not rCD93D23, was increased in a concentration-dependent manner by rCD93 proteins (Figure S1C). RAGE is a signal transduction receptor expressed in great abundance on inflammatory cells (e.g., macrophages) [18,23,24] and mediates the proinflammatory effects of HMGB1 that lead to cellular activation [16,[19][20][21][22]. Whether rCD93-HMGB1 binding affects HMGB1 interaction with RAGE and macrophages was explored. In a competitive binding assay with immobilized RAGE, adding rCD93D1 and rCD93D123 decreased HMGB1-RAGE binding in a dose-dependent manner, whereas adding rCD93D23 did not ( Figure 1D). Soluble RAGE (sRAGE), containing the extracellular ligand-binding domain of RAGE [17], can serve as a decoy by binding HMGB1 and preventing its interaction with cell surface receptors [17,24]. In flow cytometry analysis, HMGB1 can bind to the RAW264.7 macrophage surface, and HMGB1 interaction with macrophages was inhibited by sRAGE ( Figure S1D). Notably, HMGB1-macrophage interaction was inhibited by rCD93D1 and rCD93D123 but not rCD93D23. Flow cytometry analysis using fixed cells was performed to differentiate the possible effects between endocytic removal of rCD93-HMGB1 complexes and competition of rCD93 proteins with RAGE for HMGB1 binding. HMGB1 did not bind to the non-RAGEexpressing Chinese hamster ovary (CHO) cells ( Figure S2A). In contrast, HMGB1 can bind to RAW264.7 macrophages, and HMGB1-macrophage interaction was inhibited by rCD93D1 and rCD93D123 but not rCD93D23 (Figure S2B), consistent with flow cytometry analysis using live cells. These observations suggested that rCD93D1 and rCD93D123 intercepted the interaction of HMGB1 with macrophages, on which RAGE is expressed.
Taken together, these protein binding studies suggested that soluble rCD93 can bind to HMGB1 and prevent its engagement with RAGE and macrophages, and their interaction may occur at the level of D1.

Soluble rCD93 inhibits HMGB1-induced macrophage and vascular smooth muscle cell (VSMC) activation
During AAA formation, inflammatory macrophages and VSMCs are primary sources of proinflammatory cytokines (e.g., TNF-α, interleukin [IL]-6 and monocyte chemoattractant protein [MCP]-1) and extracellular matrix (ECM)-degrading proteinases (e.g., matrix metalloproteinases [MMPs]) [25][26][27]. The HMGB1-receptor interaction activates macrophage and VSMC inflammation that underpins aneurysm growth [23,28,29]. Given that soluble rCD93 interacts with HMGB1, we explored whether rCD93D123 functionally interferes with cellular activation by HMGB1. HMGB1 enhanced TNF-α secretion in PMA-differentiated THP-1 cells (Figure  2A), and this increase was reduced by rCD93D123 treatment in a dose-dependent manner. Similar results were observed in the measurements of IL-6 and MCP-1. As shown by gelatin zymography ( Figure  2B), HMGB1 increased MMP-9 and MMP-2 activities, and the increases were attenuated by rCD93D123 in a dose-dependent manner. Whether soluble rCD93 hampers the switch of VSMCs to a proinflammatory phenotype by HMGB1 was also explored. HMGB1 increased the production of TNF-α, IL-6, and MCP-1 in human aortic smooth muscle cells (HASMCs; Figure 2C), and the increase was dose-dependently inhibited by rCD93D123. Gelatin zymography revealed increased MMP-9 and MMP-2 activities by HMGB1 (Figure 2D), and the increase in MMP-9 was attenuated by 0.16 nM and 1.6 nM of rCD93D123, whereas the increase in MMP-2 was attenuated by rCD93D123 in a dose-dependent manner. Immunofluorescence staining revealed that rCD93D123 reversed morphological changes induced by HMGB1 in HASMCs (Figure 2E), as quantitatively expressed by the aspect ratio (calculated by dividing the major axis by the minor axis of the bounding ellipse for each cell [30]). The differentiated VSMC marker SM22α gene expression was suppressed by HMGB1, and the reduction was attenuated by rCD93D123 ( Figure 2F), suggesting that morphological alterations corresponded to phenotypic properties of VSMCs [31]. These findings suggested that soluble rCD93 can not only bind to HMGB1 but also functionally prevent HMGB1-induced cellular activation in macrophages and VSMCs.
Analysis of intracellular signaling showed that rCD93D123 and rCD93D1 suppressed HMGB1induced phosphorylated (p)-p65 NF-κB and p-p38 mitogen-activated protein kinase (MAPK) upregulation ( Figure 3C), and the expression levels did not significantly differ under rCD93D123 and rCD93D1. In HASMCs, rCD93D123 and rCD93D1 attenuated HMGB1-induced increases in proinflammatory cytokine secretion (Figure S3A), MMP activities ( Figure  S3B), and p-p65 NF-κB and p-p38 MAPK signaling ( Figure S3C), and the levels were similar under rCD93D123 and rCD93D1. The comparable results obtained from rCD93D123 and rCD93D1 treatment indicated that D1 of soluble rCD93 may be the effective domain impeding the proinflammatory effects of HMGB1. Consistent with protein binding studies showing that D1 of soluble rCD93 binds to HMGB1, these observations supported the concept that the proinflammatory effects of HMGB1 may be intercepted in the presence of D1 of soluble rCD93.

Treatment with soluble rCD93 attenuates AAA in vivo
Inhibition of HMGB1-RAGE interaction may suppress AAA formation [23,28]. In vivo effects of soluble rCD93 treatment were thus evaluated using two AAA models. In CaCl2-induced AAA formation, aortic dilatation was retarded in the rCD93D123treated mice (0.57 ± 0.03 mm) and the rCD93D1treated mice (0.60 ± 0.03 mm) compared with the phosphate-buffered saline (PBS)-treated mice (0.81 ± 0.04 mm, both P < 0.001; n=12 per group; Figure 4A), and the aortic diameter in the rCD93D123-treated mice was not different from that in the rCD93D1treated mice. HMGB1-RAGE interaction is crucial in maintaining chronic inflammation [18][19][20]22]. Proinflammatory cytokines (e.g., TNF-α) may activate inflammatory cells (e.g., macrophages) to induce HMGB1 secretion, and in turn, HMGB1 may induce proinflammatory cytokine release in inflammatory cells [18,19,32]. Analysis of aortic samples revealed that the high abundance of HMGB1 and RAGE ( Figure 4B) and high levels of TNF-α, IL-6, and MCP-1 ( Figure 4C) in the PBS-treated mice were substantially reduced by rCD93D123 and rCD93D1 treatment, and the levels in the rCD93D123-treated mice and the rCD93D1-treated mice were not different. Significantly fewer infiltrating monocyte/ macrophage marker antibody (MOMA)-2-positive macrophages were found in the rCD93D123-treated mice (6.2 ± 1.8 per high power field [HPF]) and the rCD93D1-treated mice (6.3 ± 1.4 per HPF) than in the PBS-treated mice (17.8 ± 2.5 per HPF; both P < 0.01; Figure 4D), and the macrophage numbers in these two groups did not differ significantly. As shown by in situ zymography, MMP activities were markedly reduced in the rCD93D123-treated and the rCD93D1-treated mice compared with the PBS-treated mice ( Figure 4E). The histological studies demonstrated loss of the natural waviness accompanied by elastin fragmentation in the PBS-treated mice, and medial elastin integrity was preserved by rCD93D123 and rCD93D1 treatment ( Figure 4F), compatible with the numbers of elastin breaks. In AngII-infused AAA, both rCD93D123 and rCD93D1 treatment effectively suppressed aortic dilatation ( Figure S4A) and increases in HMGB1, RAGE ( Figure S4B), proinflammatory cytokines (Figure S4C), infiltrating macrophage numbers (Figure S4D), MMP activities ( Figure S4E), and elastin breaks ( Figure S4F) induced by AngII infusion. The numbers/levels in the rCD93D123-and rCD93D1-treated mice were not different. The comparable effectiveness of rCD93D123 and rCD93D1 treatment revealed that D1 of soluble rCD93 may be the effective domain in suppressing AAA formation. These in vitro and in vivo findings suggested that treatment with soluble rCD93 containing D1 can attenuate inflammation and proteolysis in AAA, at least in part, through inhibition of HMGB1-RAGE signaling.
Evaluating the subcellular distribution of HMGB1 in RAW264.7 macrophages revealed that RANKL stimulated HMGB1 translocation from the nucleus to the cytoplasmic compartment and release into the culture medium in a dose-dependent manner ( Figure  S5A). Immunofluorescence staining showed that RANKL stimulation substantially accentuated the staining of HMGB1 that was colocalized with F-actin ( Figure S5B), a prerequisite cytoplasmic structure for osteoclast bone resorption [34]. Thus, RANKL may stimulate HMGB1 translocation and release in macrophages during osteoclastogenesis.
Exploring the effect of HMGB1 in the presence of RANKL on RAW264.7 macrophages using immunofluorescence staining showed that RANKL-induced F-actin ring formation was dose-dependently expanded by HMGB1 ( Figure 5A). Additionally, the area positive for the osteoclast marker tartrateresistant acid phosphatase (TRAP; Figure 5B) and TRAP activity ( Figure 5C) in RAW264.7 macrophages were increased by RANKL, and these effects were additionally enhanced by HMGB1 in a dose-dependent manner. The mRNA expression levels of osteoclast fusion genes, DCSTAMP and OCSTAMP ( Figure 5D) [37], and osteoclast differentiation marker genes, ATP6V0D2, CTSK, and TRAP (Figure 5E), were induced by RANKL. These gene expression levels were enhanced by HMGB1. These findings suggested that HMGB1 may be actively secreted under RANKL stimulation and in return promote RANKL-induced osteoclastogenesis, consistent with previous studies identifying HMGB1 as an osteoclastogenic cytokine [34].

Treatment with soluble rCD93 alleviates OVX-induced osteoporosis in vivo
Finally, the effect of rCD93D123 and rCD93D1 treatment on in vivo bone loss was evaluated using the OVX-induced model. At 8 weeks, micro-computed tomography (μ-CT) scans showed that the reduced trabecular bone volume fraction (i.e., trabecular bone volume [BV]/ total bone volume [TV]) and bone mineral density (BMD) in the PBS-treated mice were rescued in the rCD93D123-and rCD93D1-treated mice (Figure 8A), and the levels in the rCD93D123and rCD93D1-treated mice were not different. TRAP staining on histological sections showed that the increased ratios of osteoclast-covered surface (Oc.s) and eroded surface (ES) relative to the total bone surface (BS) in the PBS-treated mice were alleviated in the rCD93D123-and rCD93D1-treated mice ( Figure   8B), and the ratios in the rCD93D123-and rCD93D1treated mice were not different. Consistent with the histological findings, the increased serum level of the bone resorption marker C-terminal telopeptide of type I collagen (CTx-1) [38] in the PBS-treated mice was reduced in the rCD93D123-and rCD93D1-treated mice (Figure 8C), and the levels in the rCD93D123and rCD93D1-treated mice were not different. Taken together, the in vitro and in vivo findings suggested that treatment with soluble rCD93 containing D1 attenuated osteoclastogenesis and bone resorption at least partly through inhibiting HMGB1-receptor interaction.

Discussion
Through protein binding studies, we find that soluble rCD93 can bind to HMGB1 to prevent HMGB1 engagement with its cell receptor RAGE, and this interaction occurs at the level of D1. Accordingly, we evaluate the effect of soluble rCD93 containing D1 using a variety of in vitro and in vivo inflammatory responses, including macrophage and VSMC activation, CaCl2-induced and AngII-infused AAA formation, RANKL-induced osteoclastogenesis, and OVX-induced osteoporosis. Soluble rCD93 containing D1 ameliorates experimental AAA formation, possibly through inhibition of HMGB1-induced macrophage and VSMC activation. Also, soluble rCD93 containing D1 attenuates bone resorption, possibly through inhibition of HMGB1-enhanced osteoclastogenesis. Current knowledge regarding whether sCD93 exhibits proinflammatory or anti-inflammatory activity comes from in vitro studies, but their findings are conflicting. One study showed that soluble rCD93 induces the differentiation of monocytes to macrophage-like cells and enhances toll-like receptor (TLR) responses to lipopolysaccharide stimulation [39]. More recently, however, another group observed that soluble rCD93D1 mitigates lipopolysaccharide-induced proinflammatory responses [40]. Based on the observations obtained from different cell culture experiments and disease models, the present study demonstrates the therapeutic potential of soluble rCD93 containing D1 in inflammatory diseases and provides evidence to support the anti-inflammatory property of sCD93. In addition, the anti-inflammatory mechanism identified in the present study is quite distinct from the well-known function of sCD93 in efferocytosis. and p-p38/p38 levels at 1 hour (n = 6). ***P < 0.0001 vs. RANKL-negative group; ### P < 0.0001 vs. RANKL-only group; n.s. P > 0.9999 vs. D123-treated group. (C) Representative microscopic images of immunofluorescence staining of F-actin and quantification of F-actin size at 4 days (n = 6). ***P < 0.0001 vs. RANKL-negative group; ### P < 0.0001 vs. RANKL-only group; n.s. P > 0.9999 vs. D123-treated group. (D) Representative microscopic images of TRAP staining and quantification of TRAP-positive osteoclast area at 4 days (n = 6). ***P < 0.0001 vs. RANKL-negative group; ### P < 0.0001 vs. RANKL-only group; n.s. P > 0.9999 vs. D123-treated group. (D) TRAP activity at 4 days (n = 6). * ***P < 0.0001 vs. RANKL-negative group; ### P < 0.0001 vs. RANKL-only group; n.s. P > 0.9999 vs. D123-treated group. (E) DCSTAMP and OCSTAMP gene expression at 2 days (n = 6). ***P < 0.0001 vs. RANKL-negative group; ### P < 0.0001 vs. RANKL-only group; n.s. P > 0.9999 vs. D123-treated group. (F) ATP6V0D2, CTSK, and TRAP gene expression at 2 days (n = 6). ***P < 0.0001 vs. RANKL-negative group; ### P < 0.0001 vs. RANKL-only group; n.s. P > 0.9999 vs. D123-treated group. Yellow scale bars represent 200 μm. Black scale bars represent 1 mm. D123 indicates rCD93D123. D1 indicates rCD93D1. Data are represented as mean values ± SEM and comparative statistical analyses were done by one-way ANOVA followed by multiple comparisons. The significance of HMGB1-RAGE signaling has been established in the mechanisms of sepsis, lung injury, arthritis, and other acute and chronic inflammatory diseases [16,19,22]. The proinflammatory HMGB1 may contribute through its cell receptors, together with other DAMPs [16,[18][19][20][21][22], to a biologic amplification mechanism to drive inflammatory disorders such as aneurysm formation and bone resorption. Studies have demonstrated the roles of HMGB1 and its receptors RAGE and TLR4 in the pathogenesis of AAA [23,24,28,29]. HMGB1 neutralization using therapeutic antibody reduces proinflammatory cytokine production, attenuates macrophage infiltration and suppresses CaCl2induced AAA formation in mice [28]. RAGE deficiency prevents aortic dilatation in AngII-infused AAA formation with eliminated MMP-9 production in macrophages [24]. During aneurysm formation, HMGB1 may sustain inflammation through RAGE and TLR4 [23,29], leading to the upregulation of proinflammatory cytokines and proteinases that degrade ECM. Previous studies have also revealed the essential role of HMGB1-RAGE interaction during osteoclastogenesis [16,33,34]. RAGE-deficient mice have decreased bone resorptive activity and increased bone mass and mineral density [33]. As shown in our study, RANKL stimulated HMGB1 secretion during osteoclastogenesis. HMGB1 positively regulates RANKL-induced osteoclastogenesis in a manner dependent on RAGE [34], and sustained RAGE activation plays a vital role in chronic cellular activation and tissue damage [18][19][20]22]. RANKL collaborates reciprocally with HMGB1-RAGE signaling in inflammatory bone loss. HMGB1-RAGE interaction plays a critical role in the development of AAA and osteoporosis. Previously, extensive studies have demonstrated that the proinflammatory HMGB1 signaling sustains positive feedback loops through an autocrine or paracrine manner [18][19][20]22]. During AAA formation, the inflammatory responses mediated by HMGB1-RAGE interaction can be mitigated by soluble rCD93 containing D1, as evidenced by the reduced abundance of HMGB1, RAGE, and other inflammatory mediators. These findings are consistent with protein binding and in vitro studies showing that soluble rCD93 containing D1 can not only intercept HMGB1-RAGE interaction but also functionally inhibit HMGB1-induced macrophage and VSMC activation. In addition, the inflammatory bone loss associated with HMGB1-RAGE interaction can be alleviated by soluble rCD93 containing D1, as reflected by the reduction of HMGB1 and RAGE abundance and osteoclastogenic marker gene expression during osteoclastogenesis. These findings, together with protein binding study results, are in line with the observations in AAA and further corroborate the concept that soluble rCD93 containing D1 sequesters HMGB1 to ameliorate inflammatory diseases.
Interfering with HMGB1-receptor interaction by HMGB1 sequestration is proposed as a promising strategy to alleviate HMGB1-related inflammatory disorders [16,21]. As demonstrated in the present study, soluble rCD93 ectodomain proteins containing D1 can bind to HMGB1. The therapeutic benefits of rCD93D123 and rCD93D1 in HMGB1-driven inflammatory diseases suggest that soluble rCD93 containing D1 acts like a decoy receptor. Such a strategy can also be accomplished by several established approaches [23,28,34,35], such as HMGB1 neutralizing antibody, sRAGE, and recombinant thrombomodulin (CD141) containing the lectin-like domain.
Among the C-type lectin-like domain group 14 family members, including thrombomodulin, tumor endothelial marker 1 (TEM1; CD248, also called endosialin), and C-type lectin domain containing 14A, the amino acid alignment of human CD93 is most closely related to that of thrombomodulin [2,40]. It is speculated that CD93 and thrombomodulin are derived from a common ancestor due to their strong homology and close proximity on chromosome 20 [2,3,40,41]. The lectin-like domain of soluble form thrombomodulin can regulate inflammation by sequestering HMGB1 and lipopolysaccharide [17,42]. Through these actions, soluble recombinant thrombomodulin containing the lectin-like domain exerts therapeutic benefits in a variety of inflammatory disease models in vivo [17,35,[42][43][44]. In the present study, the observations from in vitro and in vivo experiments suggest the anti-inflammatory function of soluble rCD93 containing D1, the lectin-like domain. These findings provide evidence to support the long-standing speculation of functional similarity in modulating inflammation between CD93 and thrombomodulin [2]. Correspondingly, given that membrane-bound thrombomodulin (i.e., thrombomodulin expression on the cell surface) in macrophages exerts a proinflammatory function that is opposite to the general understanding regarding the anti-inflammatory activity of soluble thrombomodulin [45,46], the role of membrane-bound CD93 during inflammation needs to be reappraised. Interestingly, current understanding regarding the function of membrane-bound CD93 in inflammatory responses is also contradictory [47][48][49][50]. CD93 deficiency leads to increased inflammation and severity in experimentally-induced cerebral ischemia and encephalomyelitis [48,49], consistent with the finding that CD93-deficient mice have increased leukocyte recruitment during thioglycollate-induced peritonitis [47]. However, a recent study demonstrated that membrane-bound CD93 might act as a cell surface receptor for exogenous DNA and mediate downstream inflammation in a CD93-expressing IMR-32 neuroblastoma cell line [50], implying a proinflammatory property of membrane-bound CD93. Further studies are warranted to explore whether membrane-bound CD93 exerts a proinflammatory or anti-inflammatory function in inflammatory responses.

Conclusions
In conclusion, the present study demonstrates an anti-inflammatory property of sCD93/rCD93 residing in the lectin-like domain. These findings provide insight by demonstrating a novel anti-inflammatory mechanism, i.e., sequestering HMGB1 by sCD93 via the lectin-like domain, distinct from the function of sCD93 in efferocytosis. Links between sCD93 containing the lectin-like domain and the protective effect against HMGB1 suggest that sCD93-HMGB1 interaction may be relevant to in vitro and in vivo inflammatory responses and that the soluble rCD93 lectin-like domain holds promise in potential therapies for AAA and osteoporosis.
Assays for rCD93-HMGB1 interaction. For co-immunoprecipitation (Co-IP) assay, rCD93D123 (5 nM) and HMGB1 (5 nM; R&D Systems, Minneapolis, MN) were incubated for 1 hour at 37 °C in immunoprecipitation buffer (50 mM Tris-HCl, 100 mM NaCl, 0.25% Triton X-100, 1 mM MgCl2, and 2 mM EGTA; final volume: 0.5 ml). Protein A agarose (20 µl; Millipore, Billerica, MA) conjugated with a rabbit polyclonal antibody against HMGB1 (1 µg; ab18256, Abcam, Cambridge, MA) or normal IgG, serving as a negative control, was added to the protein mixtures (200 µl) and then incubated overnight at 4 °C. Protein A agarose was collected by centrifugation and washed. The immune precipitate was solubilized in sample buffer (10 µl). Samples were loaded to 10% SDS-PAGE followed by a western blot assay with anti-c-Myc antibody (sc-40, Santa Cruz Biotechnology; 1:1000). Subsequently, the membranes were incubated with the appropriate secondary antibodies at room temperature for 1 hour. The immunoreactive bands were detected by chemiluminescence reagents (Millipore). Far-western blot assay was conducted as the following. rCD93D1, rCD93D23, rCD93D123, and bovine serum albumin (BSA; 0.1 µg/ml) were separated by 10% SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and then hybridized with HMGB1 (0.5 µg/ml) in PBS overnight at 4 °C. After appropriate wash steps, the membranes were incubated with rabbit anti-HMGB1 antibody (ab18256, Abcam; 1:1000) overnight at 4 °C, followed by an HRP-conjugated secondary antibody. Ponceau S staining (Sigma-Aldrich, St Louis, MO), serving as the loading control, was performed based on the manufacturer's protocols. ELISA solid-phase binding assay was performed as the following. Ninety-six-well ELISA plates with high proteinbinding capacity (Nunc, Roskilde, Denmark) were coated with indicated doses of rCD93D123, rCD93D1, rCD93D23, and BSA overnight at 4 °C with coating buffer (0.15% Na2CO3, 1% MgCl2, and 0.3% NaHCO3 in double-distilled water). The wells were washed with phosphate-buffered saline with Tween 20 (PBST) and then incubated with blocking buffer (50 mg/ml BSA, 20 mM Tris-HCl, 0.15 M NaCl, and 5 mM CaCl2 in double-distilled water) for 2 hours at room temperature. After PBST washes, indicated doses of HMGB1 were incubated with reaction buffer (1 mg/ml BSA, 20 mM Tris-HCl, 0.15 M NaCl, and 5 mM CaCl2 in double-distilled H2O) overnight at 4 °C. The competitive ELISA assay was performed as the following. Ninety-six-well ELISA plates were coated with sRAGE (10 nM; Sino Biological, Beijing, China) overnight at 4 °C with a coating buffer. After being blocked with a blocking buffer and washed with PBST, the indicated concentration of rCD93D123, rCD93D1, rCD93D23, HMGB1, and BSA were incubated in the sRAGE-coated plate with reaction buffer overnight at 4 °C. The ELISA plates were washed and then incubated with anti-HMGB1 antibody (ab18256, Abcam; 1:1000) for 2 hours at room temperature, followed by an HRP-conjugated antibody (GeneTex, Irvine, CA) for 2 hours at room temperature. Tetramethylbenzidine (Sigma-Aldrich) was applied as a substrate and incubated for 20 minutes, followed by the addition of 50 μl of 1 M H2SO4 per well to stop the reaction. The absorbance at 450 nm was measured in a microplate reader (DYNEX Technologies, Denkendorf, Germany). For flow cytometry assay, live or fixed RAW264.7 and CHO cells were incubated with HMGB1 (20 nM) in the presence of sRAGE, rCD93D123, rCD93D1, or rCD93D23 (10 nM each) for 1 hour at 37 °C. Cells were stained with anti-HMGB1 antibody (ab18256, Abcam; 1:100) overnight at 4 °C followed by Alexa Fluor 546-conjugated rabbit IgG (Invitrogen, Carlsbad, CA; 1:500) for 30 minutes at room temperature and analyzed by fluorescence activated cell sorting (FACS; BD Biosciences, San Jose, CA). The histogram was generated using WinMDI 2.9 software (Purdue University cytometry laboratories, West Lafayette, IN).
Regarding potential contamination in the HMGB1 preparation, according to the datasheet, the lipopolysaccharide level was less than 0.1 EU per 1 μg of the protein measured by the Limulus amoebocyte lysate assay. The amount of DNA in HMGB1 (100 ng) was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) with a negative control (Milli-Q ultrapure water) and a positive control (genomic DNA isolated from the mouse tail). The results suggested the absence of DNA in the HMGB1 preparation (Table  S1).
In vitro cell culture. Human monocytic THP-1 cells were obtained from the Bioresource Collection and Research Center in Taiwan. Cells were maintained in an RPMI medium with 10% FBS, 1 mM L-glutamine, sodium pyruvate, and penicillin/streptomycin in a 5% CO2 incubator at 37 °C and were plated in a medium containing 10 nM phorbol-12myristate-13-acetate (PMA; Sigma-Aldrich) for 18 hours for differentiation. HASMCs (Cascade Biologics, Portland, OR) were cultured in Medium 231 containing 10% FBS, 1% smooth muscle growth supplement, and penicillin/streptomycin in a 5% CO2 incubator at 37 °C. Mouse RAW264.7 cells (Bioresource Collection and Research Center) were cultured in α-MEM with 1% FBS, 2 mM L-glutamine, and penicillin/streptomycin within a 5% CO2 incubator at 37 °C. In addition to, RAW264.7 cells were stimulated with macrophage-colony stimulating factor (20 ng/ml) and indicated doses of RANKL for osteoclast differentiation. The culture medium was replenished every 2 days. THP-1 cells, HASMCs, and RAW264.7 cells were treated with HMGB1, rCD93D123, and rCD93D1 as indicated in the experiments.
AAA models and soluble rCD93 treatment. The CaCl2-induced and AngII-infused AAA mouse models were established as previously described [44,51]. These mice were maintained in a pathogen-free animal facility at the Animal Center of National Cheng Kung University. Only male mice were used due to the potential influence of female sex hormones on AAA models and the male predominance of human AAAs [52,53]. For CaCl2-induced AAA, the male C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME) aged 9-12 weeks were used. The CaCl2-induced AAA model was induced with 0.5 M CaCl2 applied onto the infrarenal aorta for 15 minutes under anesthesia. For AngII-infused AAA, the male ApoE -/mice (The Jackson Laboratory) aged 6 months were used. The AngII-infused AAA model was generated using the continuous subcutaneous infusion of AngII (1000 ng/kg/min; Sigma-Aldrich) via osmotic pumps ( OVX-induced osteoporosis model and soluble rCD93 treatment. Female C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME) aged 8 weeks were used and maintained in a pathogen-free animal facility at Kaohsiung Medical University Animal Center. The OVX-induced bone loss model was generated by bilateral ovariectomy under anesthesia [54]. To evaluate the effectiveness of soluble rCD93 treatment in OVX-induced bone loss, rCD93D123 (intraperitoneal injection with rCD93D123 [0.6 mg/kg] in 0.1 mL PBS once a day every 3 days since the following day after OVX), a mole-equivalent dosage of rCD93D1 (intraperitoneal injection with rCD93D1 [0.2 mg/kg] in 0.1 mL PBS once a day every 3 days since the following day after OVX), or the same volume of PBS (n = 12 per group) was used. At 8 weeks, the mice were sacrificed for μ-CT evaluation and lower limb harvest. All μ-CT analyses were performed in accordance with the guidelines [55]. Bone samples from each group were imaged using a SkyScan-1076 μ-CT System (Skyscan, Kontich, Belgium). The μ-CT scanner was operated at 45 kV, 220 μA, 0.4 μ rotation step, 0.5 mm aluminum filter, and a scan resolution of 18 μm/pixel. The following 3D parameters, including BMD, TV, BV, and BV/TV, were evaluated using CT Analyzer software (Bruker, Kontich, Belgium). These experiments were approved by the Animal Care and Use Committee of Kaohsiung Medical University Animal Center (approval number: 110004) and conformed to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health (NIH Publication #85- 23, revised 1996).
In vitro immunofluorescence assay. After stimulation, HASMCs and RAW264.7 cells were washed with PBS and fixed with 4% paraformaldehyde at room temperature for 30 minutes, followed by permeabilization with 0.1% Triton X-100 for 5 minutes at 4 °C. After being blocked with PBS containing 10% BSA for 1 hour at room temperature, cells were incubated with anti-SM22α (ab14106, Abcam; 1:500) or anti-HMGB1 (ab18256, Abcam; 1:500) antibodies overnight at 4 °C. Cells were washed and incubated with fluorescent-dye conjugated secondary antibodies (Invitrogen) at room temperature for 1 hour, washed using PBS, and then stained with 4', 6-diamidino-2phenylindole (DAPI) for 5 minutes. The fluorescence was visualized using a confocal laser scanning microscope coupled with an image analysis system (Olympus Fluoview FV1000, Olympus, Tokyo, Japan). For the morphological changes of HASMCs, 3 random view-fields were used to measure the aspect ratio of individual cells. The aspect ratio was calculated through the division of the major axis by the minor axis of the bounding ellipse for each cell using Imaris software (Andor Technology, Belfast, Northern Ireland) [30,56].
Western blot analysis and gelatin zymography. Cells were harvested using ice-cold lysis buffer after being washed twice with ice-cold PBS. The protein homogenate from cells or aortic samples was centrifuged at 12000 rpm for 25 minutes at 4 °C, and the supernatant was recovered as the total cellular protein. Separation and preparation of cytoplasmic and nuclear extracts were performed using the NE-PER Nuclear and Cytoplasmic Extraction Reagent Kit (Pierce Biotechnology, Rockford, IL) as needed according to the manufacturer's instructions. The protein from each sample was separated by SDS-PAGE on 12% acrylamide gels, transferred to a PVDF membrane, and then blocked with 5% non-fat dry milk in Tris-buffered saline with Tween 20. The membrane was incubated with primary antibodies in PBST at 4 °C overnight. The primary antibodies used in the study were antibodies against human and mouse p-p65 (#3033, Cell Signaling, Danvers, MA; 1:1000), p65 (sc-8008, Santa Cruz; 1:1000), p-p38 (#4511, Cell Signaling; 1:1000), p38 (#9212, Cell Signaling; 1:1000), HMGB1 (ab18256, Abcam; 1:1000), RAGE (A13264, ABclonal, Woburn, MA,; 1:1000), GAPDH (GTX627408, GeneTex; 1:1000), α-tubulin (GTX112141, GeneTex; 1:1000), and lamin B2 (33-2100, Thermo Fisher, Waltham, MA; 1:1000). Subsequently, the membranes were incubated with appropriate secondary antibodies at room temperature for 1 hour, followed by chemiluminescence reagents (Millipore). MMP-2 and MMP-9 activities in the supernatants obtained from the culture medium were demonstrated using gelatin zymography. The supernatant was separated using 7.5% SDS-PAGE containing 0.1% gelatin. After incubation with the activation buffer overnight at 37 °C, SDS-PAGE was performed with Coomassie blue staining. The band intensity in western blot analysis and gelatin zymography was quantitatively measured using ImageJ software. GADPH, α-tubulin and lamin B2 levels were determined to confirm equal protein loading.
TRAP activity, TRAP staining, and F-actin ring formation in cell culture. TRAP activity was measured using 4-nitrophenyl phosphate (4-NPP) as a substrate [35]. A 50-μl cell lysate was incubated with 100 μl of substrate solution (0.1 M 4-NPP, 0.1 M sodium acetate, and 0.2 M sodium tartrate, pH 5.0) for 60 minutes at 37 °C. The reaction was terminated by adding 50 μl of 3 M NaOH. Absorbance was measured at 405 nm in a microplate reader (DYNEX Technologies). For TRAP staining, cells were fixed and stained with a TRAP staining kit (387A, Sigma-Aldrich). TRAP-positive multinucleated cells with 3 or more nuclei were counted as osteoclasts. The TRAP-positive osteoclast area per well was quantified using AxioVision software (Carl Zeiss, Oberkochen, Germany). To observe F-actin ring formation, cells were fixed with 4% paraformaldehyde for 30 minutes and permeabilized with 0.1% Triton X-100 for 5 minutes. After being blocked with PBS containing 10% BSA for 1 hour, cells were incubated with rhodamine-conjugated phalloidin (Invitrogen; 1:1000) in PBS for 1 hour at room temperature, followed by Supplementary figures and table. https://www.thno.org/v13p4059s1.pdf