C5aR1 interacts with TLR2 in osteoblasts and stimulates the osteoclast‐inducing chemokine CXCL10

Abstract The anaphylatoxin C5a is generated upon activation of the complement system, a crucial arm of innate immunity. C5a mediates proinflammatory actions via the C5a receptor C5aR1 and thereby promotes host defence, but also modulates tissue homeostasis. There is evidence that the C5a/C5aR1 axis is critically involved both in physiological bone turnover and in inflammatory conditions affecting bone, including osteoarthritis, periodontitis, and bone fractures. C5a induces the migration and secretion of proinflammatory cytokines of osteoblasts. However, the underlying mechanisms remain elusive. Therefore, in this study we aimed to determine C5a‐mediated downstream signalling in osteoblasts. Using a whole‐genome microarray approach, we demonstrate that C5a activates mitogen‐activated protein kinases (MAPKs) and regulates the expression of genes involved in pathways related to insulin, transforming growth factor‐β and the activator protein‐1 transcription factor. Interestingly, using coimmunoprecipitation, we found an interaction between C5aR1 and Toll‐like receptor 2 (TLR2) in osteoblasts. The C5aR1‐ and TLR2‐signalling pathways converge on the activation of p38 MAPK and the generation of C‐X‐C motif chemokine 10, which functions, among others, as an osteoclastogenic factor. In conclusion, C5a‐stimulated osteoblasts might modulate osteoclast activity and contribute to immunomodulation in inflammatory bone disorders.

indicating an important function in these cells. 5,6 Indeed, in osteoblasts C5a induces migration and expression of the inflammatory cytokines interleukin-6 (IL-6) and IL-8, and receptor activator of nuclear factor kappa B ligand (RANKL), which is essential for osteoclast formation and activity. 5,7 Moreover, C5aR1-knockout (-ko) mice display reduced osteoclast numbers and significantly increased bone mass, suggesting that C5a/C5aR1 signalling might regulate physiological bone turnover. 8 The C5a/C5aR1 axis in bone cells might be particularly relevant under pathological conditions, because mice lacking C5aR1 are protected against arthritis, 9 and C5aR1 activity has been linked to substantial bone loss in a periodontitis model. 10 Antagonizing C5aR1 significantly reduced periodontal inflammation and subsequent bone loss in this model. 11 Moreover, we previously demonstrated that C5aR1 was strongly expressed in osteoblasts in response to bone injury, 7 and that bone fracture healing in a rodent model of severe systemic inflammation significantly improved when treated with a small peptide C5aR1 antagonist. 12 In this setting, osteoblasts were found to be target cells for C5a, because mice with an osteoblast-specific C5aR1 overexpression displayed impaired fracture healing. 6 However, the molecular mechanisms underlying the C5a/C5aR1 signalling axis in osteoblasts remain unclear, also in respect of potential cross-talking signalling pathways, which can modulate or are modulated by C5aR1 actions. In immune cells, C5aR1 has been described to interact with other immune receptors, including receptors for immunoglobulin G (IgG) antibodies, the FcγRs, 13 or with other biological systems, including the coagulation cascade. 14,15 Tolllike receptors (TLRs) are further potential interaction candidates, because, similar to the complement system and its receptors, they are important for early recognition and adequate response to danger molecules. 16 In this regard, pathways downstream of complement receptors and TLRs interact in various immune cells, 17,18 thereby modulating inflammatory responses. 19 In this study, we aimed to determine the intracellular events following C5aR1 activation in osteoblasts. We analysed gene expression patterns and intracellular signalling pathways upon C5aR1 activation and found a strong modulation of genes involved for example in the mitogen-activated protein kinase (MAPK) and insulin pathways. Furthermore, we demonstrated that C5aR1 and TLR2 interact in osteoblasts, resulting in upregulation of the immune cell chemoattractant C-X-C motif chemokine 10 (CXCL10), which can induce osteoclastic bone resorption. [20][21][22] These results suggest that complement-activated osteoblasts are able to modulate the inflammatory milieu during inflammatory bone diseases in concert with osteoclasts and immune cells.

| Mouse model
Male wild-type (WT) control (C57BL/6) mice were purchased from Charles River (Sulzfeld, Germany) while C5aR1-ko mice, originally generated by C. Gerard 23 and kindly provided by John D. Lambris (University of Pennsylvania, USA), were bred in-house. Mice were housed according to the guidelines for the care and use of laboratory animals (ARRIVE) and had access to a standard mouse feed (ss-niffR R/M-H, V1535-300, Ssniff, Soest, Germany) and water ad libitum. Experiments were performed with permission of the local authorities.

| Osteoblast isolation, cultivation, and stimulation
Primary osteoblasts were isolated from long bones of 8-12-week-old mice and differentiated for 14 days, as described previously. 6,24 Briefly, harvested diaphyses were shred and digested for 2 h using 125 U/ml collagenase type II (Sigma-Aldrich, Steinheim, Germany) in

| Enzyme-linked immunosorbent assay (ELISA)
Cell-culture supernatants of osteoblasts stimulated for 4 h with C5a and/or Pam3, were analysed according to the manufacturer's instructions using a mouse ELISA-kit for CXCL10 (CRG-2) (#EMCXCL10, Thermo Fisher Scientific). Data were analysed using a standard curve provided with the kit. Values below assay detection limit were set to zero.

| Reverse transcription-PCR (RT-PCR)
Total RNA isolation and RT-PCR were performed as described previously. 6 Gene expression was analysed relative to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) using the ΔΔCt method. Primers were purchased from Invitrogen (Thermo Fisher Scientific, Waltham, USA) and sequences are available in Table S1. Capillary gel electrophoresis and quantification of PCR products were performed using QIAxcel DNA Screening Gel Cartridge on a QIAxcel Advanced System (Qiagen, Hilden, Germany). values were calculated using the robust multiarray average. Differentially expressed probesets were determined by t test and considered statistically significant when P < 0.05 and fold change ≤ 1.5, as published previously. 25 Functional protein-association networks were identified using the STRING 10 program ( 26 http://string-db.org/).

| Microarray-based gene expression analysis
Differentially expressed genes were subjected to pairwise gene ontology (GO) term similarity measure with Lin's algorithm using GOSemSim in R. 27 Similarity matrices served as inputs for hierarchical clustering using the R package hclust. Enrichment analysis of the resulting groups was performed using EnrichR ( 28 http://amp.pharm. mssm.edu/Enrichr/). The GoMINER tool 29 was used to identify the most affected biological processes and pathway analysis was performed using Transcriptome Analysis Console (Affymetrix). Complete microarray data are available at Gene Expression Omnibus (GEO accession number: GSE107036).

| Coimmunoprecipitation
Immunoprecipitation of C5aR1 and subsequent C5aR1 and TLR2 detection by immunoblotting was performed using MC3T3-E1 cells,

| Immunoblotting
Cells were lysed in Pierce ® RIPA buffer (Thermo Fisher Scientific), containing PPi. Sample buffer was added to the protein samples and RT. WesternBright TM ECL chemiluminescent HRP substrate (Advansta, Menlo Park, USA) was applied to the membranes for 2 min at RT and the signal was captured by membrane exposure to X-ray film (CL-XPosure TM , Thermo Fisher Scientific). Western blot images were quantified by use of densitometry values derived by Adobe Photoshop CS6 and are presented relative to GAPDH.

| Immunofluorescent staining
Osteoblasts were fixed in 4% phosphate-buffered formalin and unspecific binding was prevented using goat serum for 1 h at RT.

| Statistical analysis
Results are presented as the mean ± standard deviation. For statistical analysis, the software GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, USA) was used. Testing for normal distribution was performed using the Shapiro-Wilk test. Student's t test was applied when two groups were compared while one-way analysis of variance (ANOVA), followed by Fisher's LSD post hoc test, was applied to compare three or more groups. The level of significance was set at P ≤ 0.05.

| C5a regulates genes involved in the MAPKand transforming growth factor (TGF)-β pathways, insulin signalling, and the activator protein (AP)-1 transcription factor in osteoblasts
Confirming previous findings, 5,6 increased C5ar1 expression levels were detected in osteoblasts cultured in osteogenic differentiation medium compared to cells in normal proliferation medium ( Figure 1A and B). Immunofluorescent staining demonstrated strong C5aR1 upregulation upon differentiation ( Figure 1C). To investigate actions conveyed by the C5a/C5aR1 axis, we performed gene expression profiling of C5a-stimulated osteoblasts. In total, 606 probesets were differentially regulated, which strongly clustered between the treatment groups, as visualized by heat-map (Figure 2A), and regarding their presumed protein functions, exemplified by the protein-association network ( Figure 2B). After excluding probesets not assigned to an Entrez Gene ID, 472 probesets remained for further analysis, whereof 241 probesets were upregulated ( Figure 2C) and 231 probesets were downregulated ( Figure 2E). Hierarchical clustering and enrichment analysis revealed that upregulated genes are involved in insulin signalling and secretion, TGF-β receptor signalling and the activation of MAPKs ( Figure 2D). Downregulated genes are involved in AP-1 transcription factor formation and interferon production and signalling ( Figure 2F). By performing pathway analysis of all regulated genes, insulin metabolism, TGF-β signalling, MAPK regulation and interferon signalling appeared among the top C5a-regulated biological pathways (Table 1), thus confirming data derived from the enrichment analyses ( Figure 2D and F). A list of the top 20 regulated pathways upon C5a stimulation is provided in the supplement (Table S2). To confirm microarray findings, differential expression of selected candidate genes was validated by RT-PCR.
This shows that C5a activates MAPKsignalling in osteoblasts. The AP-1 transcription factor subunits Fos and Jun (Table 2), and levels of other immediate early genes, including immediate early response 2 and 3 (Ier2, Ier3) and early growth response 2 and 3 (Egr2, Egr3) (Table S3) were considerably reduced. There was a high correlation (R 2 = 0.957) between the logarithmic fold change values derived from the microarray and RT-PCR analyses ( Figure S2). Complete lists of the top 20 upregulated (Table S3) and downregulated genes (Table S4) upon 4 h-C5a-treatment are provided in the supplement.

| C5aR1 and TLR2 interact in osteoblasts and downstream signalling involves the activation of p38 MAPK
'Toll-like Receptor Signalling' was among the top C5a-regulated pathways ( Table 1) and thus any interplay between C5aR1 and TLR2 in osteoblasts was of special interest. We first analysed TLR2 expression and detected slightly but significantly upregulated gene levels upon osteogenic differentiation ( Figure 3A   osteoblasts ( Figure 3E). This interaction was apparent already under unstimulated conditions, and increased when prestimulating C5aR1.
Stimulation of TLR2 with Pam3 did not further increase receptor interactions ( Figure 3E). We did not find evidence supporting a reciprocal regulation of the receptors, as C5aR1 gene and protein expression was unaltered after stimulation with Pam3, neither were levels of TLR2 altered upon stimulation with C5a ( Figure S2). Nevertheless, we confirmed microarray findings, that C5a treatment enhanced levels of Toll-interleukin 1 receptor domain-containing adaptor protein (Tirap) ( Figure 3F), an important molecule for TLR downstream actions. Tirap levels were increased as early as 30 min and remained high until 24 h after C5a stimulation. Therefore, C5a appears to enhance TLR2 downstream signalling rapidly and persistently ( Figure 3F). For confirmation, we investigated receptormediated activation of MAPKs, which are intracellular signalling transducers, and which we found being regulated by C5a on gene level ( Figure 2D; Table 2). MAPKs p38 and ERK1/2 (data not shown) were phosphorylated and thereby activated by C5a ( Figure 3I). The same effect was observed after TLR2 stimulation. Notably, p38 phosphorylation significantly increased more at 30 min after receptor costimulation, compared to isolated receptor stimulation ( Figure 3I), which was confirmed by the quantification of the western blot protein bands ( Figure 3J). This finding suggests combined actions of C5aR1 and TLR2 in downstream signalling. Importantly, inhibition of C5aR1 with its antagonist PMX-53 prevented p38 phosphorylation ( Figure 3G). In line, p38 phosphorylation was considerably diminished in C5aR1-ko osteoblasts ( Figure 3H), corroborating C5aR1dependent MAPK activation. Additionally, TLR2-mediated p38 phosphorylation was prevented by inhibiting the TLR adaptor protein MyD88 ( Figure 3G). Therefore, in addition to C5aR1, signalling via p38 MAPK is also TLR2-dependent.  The main regulated biological pathways, determined by using the Transcriptome Analysis Console Software, are shown in descending order, based on their significance. The number of up-and downregulated genes, included in the respective pathways according to the gene ontology term, is shown in columns 3 and 4. The P-value of differential regulation and its negative decadal logarithm (significance) are shown in column 6 and 5, respectively. A complete list of the top 20 regulated pathways is provided in the supplement (Table S1).
receptor costimulation led to significantly higher CXCL10 levels compared to isolated receptor stimulation ( Figure 4B). C-X-C motif chemokine receptor 3 (CXCR3), which is the receptor for CXCL10, was  Figure 4D and E). In contrast to the conditioned medium from C5a-and/or Pam3-treated osteoblasts, the direct addition of C5a and Pam3 to the osteoclast medium did not induce osteoclastogenesis. Furthermore, the inhibition of C5aR1 and MyD88, simultaneously to the incubation with osteoblast-conditioned medium, did not impair its osteoclastogenic potential ( Figure 4D and E). To examine the osteoclastogenic effect of CXCL10 separately, recombinant CXCL10 was added to an additional treatment group, which showed strongly enhanced osteoclast formation, while this effect was reversed using a neutralizing CXCL10-antibody ( Figure 4D). Importantly, osteoclast formation mediated by the osteoblast supernatants was significantly attenuated when antagonizing CXCL10 ( Figure 4D and E).
The effects on osteoclast formation were confirmed on gene level, as genes encoding for TRAP (Acp5) and Cathepsin K (Ctsk) were induced by C5a-and Pam3-treated osteoblast supernatant, and CXCL10 (data not shown). The findings of this study, regarding C5aR1 and TLR2 interactions in osteoblasts and their convergence in downstream signalling pathways, are illustrated in the current working model ( Figure 5).

| DISCUSSION
In this study, we demonstrated that C5a modulates the expression of genes involved in the MAPK and TGF-β pathways, insulin and interferon signalling and the AP-1 transcription factor in osteoblasts.
We further showed that C5aR1 and TLR2 interact in osteoblasts and crosstalk in downstream signalling. The pathways converge on the activation of p38 MAPK, eventually leading to expression of the chemokine CXCL10. To unravel intracellular events following activation of the C5a/C5aR1 axis, we performed whole genome-covering microarray analyses. Interestingly, we found a C5a-mediated induction of genes related to insulin signalling and glucose metabolism.
Negative insulin regulators were downregulated, while Gfpt2 expression was upregulated. GFPT2 regulates glucose flux, metabolism, and utilization. It converts fructose-6-phosphate to glucosamine-6-phosphate and thereby catalyses the rate-limiting step of the hexosamine biosynthesis pathway (HBP). High glucose flux into the HBP is associated with insulin resistance, impaired glucose tolerance and type 2 diabetes, 30 effects which could be linked to increased GFPT2 activity. 31,32 Notably, C5a influences glucose metabolism in neutrophils, leading to increased glucose uptake and glycolysis, thus resembling insulin action in these cells. 33 insulin resistance in an in vivo obesity model. 35 These data imply that complement, and in particular the C5a/C5aR1 axis, might affect glucose metabolism not only in immune cells but also in osteoblasts.
This might be the case particularly under high bone-turnover conditions, requiring increased energy supply. 36  required. C5a-activated HBP may enhance proteoglycan production, which is important for bone structure and interacts with growth factors present in bone matrix, such as TGF-β. 37,38 Interestingly, we found C5a-induced genes to be involved in the TGF-β pathway.
TGF-β is released from bone matrix during bone resorption, co-ordinating bone formation and osteoblast activity. It promotes osteoblast differentiation 39 and migration of osteoblast precursor cells to the bone-turnover site. 40  Indeed, studies showed that osteoblasts produce the osteoclast-stimulator RANKL in response to bacterial-induced TLR2 activation. 48,53 Additionally, C5aR1-mediated effects in osteoblasts involve the induction of RANKL. 5  which we confirmed here on gene level. Therefore, osteoclasts are indeed potential target cells of osteoblast-secreted CXCL10, which could in turn also activate osteoblasts in a feedback loop. The performed osteoclast formation assay did confirm the osteoclastic potential of both CXCL10 alone and C5a-and Pam3-induced osteoblast-secreted CXCL10. By using a neutralizing antibody against CXCL10, we showed that the osteoclastogenic effect mediated by the applied osteoblast conditioned media was significantly reduced, however, not completely abolished. These findings indicate that CXCL10 is a crucial, but probably not the only osteoclastogenic factor induced by C5aR1 and TLR2 activation.
A limitation of this study is that we focused on C5aR1-mediated effects and did not entirely distinguish between the effects of the two receptors for C5a, C5aR1, and C5aR2. In a recent in vivo study, we demonstrated that the lack of C5aR1 or C5aR2 differentially affected bone cells and the early inflammatory phase of fracture healing. 8 Therefore, further studies are required, dissecting C5aR1from C5aR2-mediated effects in osteoblasts, to enable tailored C5a receptor-modulation under inflammatory bone conditions in future.
In summary, we demonstrated the interaction of C5aR1 and TLR2 in osteoblasts, not only physically but also functionally regard-

CONFLI CT OF INTEREST
The authors confirm that there are no conflicts of interest.

AUTHOR CONTRI BUTIONS
YM: research design, data acquisition, data analysis and interpretation, manuscript preparation; AR: data interpretation and revision of the manuscript; JP and KH: data analysis and interpretation; AK, MHL, and MHL: data interpretation, scientific discussions; AI: