Structure-Activity Relationship of Anthocyanidins as an Inhibitory Effect on Osteoclast Differentiation

their aglycons Anthocyanidins various biological functions, such as anti-oxidant effects. their relationship in bone tissue is not known. In this study, we examined the effects of three anthocyanidins, delphinidin, cyanidin and pelargonidin, on osteoclast differentiation and bone resorption to elucidate the structure-activi-ty relationship. Anthocyanidins suppressed both IL-1 and LPS induced osteoclast differentiation in cocultures of bone marrow cells and primary osteoblasts, and bone resorbing activity in calvarial organ cultures. In osteoblasts, anthocyanidins inhibited prostaglandin (PG) E 2 production via the downregulation of membrane-bound PGE synthase (mPGES)-1, leading to the suppression of PGE 2 -mediated receptor-activator of nuclear factor-kappa B (NF-κB) ligand (RANKL) expression. In osteoclasts, anthocyanidins inhibited RANKL-induced osteoclast differentiation through the downregulation of osteoclast differentiation marker genes, nuclear factor of activated T-cells 1 (NFATc1), cathepsin K and tartrate-resistant acid phosphatase (TRAP). We further found that anthocyanidins suppressed the inhibitor of NF-κB kinase (IKK) activity in vitro assay, a signal component of NF-κB pathway, suggesting IKK was a novel target molecule of anthocyanidins. We found that delphinidin exerted the most potent inhibitory activity in these experiments, compared with cyanidin and pelargonidin. Anthocyanidins exhibits inhibitory activity in bone resorption, which may depend on the number of hydroxide residues.


INTRODUCTION
Bone remodeling is precisely regulated by osteoblastic bone formation and osteoclastic bone resorption. Osteoclasts are the bone-resorbing cells differentiated from macrophage lineage cells, and excess activation of osteoclasts induced by inflammatory molecules results in bone-related diseases including periodontitis. 1) Osteoclast differentiation is regulated by the interaction of receptor activator of nuclear factor kappa B (NF-κB) ligand (RANKL) expressed on osteoblasts and RANK expressed osteoclast precursor cells. RANKL-RANK signalling activates various molecules including NF-κB and nuclear factor of activated T cells (NFATc) 1, a master transcription factor for osteoclast differentiation and downstream component of NF-κB. 2) RANKL expression in osteoblasts is induced by various inflammatory molecules, such as interleukin (IL)-1 and lipopolysaccharide (LPS). LPS, an outer membrane component of gram-negative bacteria, is a major patho-genic molecule of the periodontitis and identified as a ligand of toll-like receptor (TLR) 4. 3) We previously reported that LPS-TLR4 signalling induced membrane-bound prostaglandin E synthase (mPGES)-1-mediated PGE 2 production via NF-κB pathway, and PGE 2 induced RANKL expression in osteoblasts, leading to inflammatory bone resorption of alveolar bone in a model of periodontitis. 4) PGE 2 is a major inflammatory mediator synthesized thorough arachidonic acid cascade in osteoblasts. In this cascade, free arachidonic acid is released from plasma membrane phospholipid by phospholipase (PL) A 2 , and cyclooxygenases (COXs) converts free arachidonic acid into PGH 2 . Subsequently, PGE 2 is synthesized from PGH 2 by PGE synthases (PGESs) and is secreted from cells. We also previously reported that PGE 2 recognized by PGE receptor (EP) 4 in osteoblasts induced RANKL expression and osteoclast differentiation. 5,6) Anthocyanins (ACNs), a member of flavonoids, are abundantly contained in reddish vegetables and fruits such as red

Regular Article
Structure-Activity Relationship of Anthocyanidins as an Inhibitory Effect on Osteoclast Differentiation cabbages, berries and red grapes. Their aglycons are called anthocyanidins (ACDs), and there are some structure variants including delphinidin, cyanidin, pelargonidin. Fig. 1A showed the structure of delphinidin, cyanidin and pelargonidin having 3, 2 and 1 hydroxyl groups on B ring, respectively. ACDs have been reported to exhibit anti-inflammatory and anti-oxidative effects, and the strength of their activities could relate to their structure. [7][8][9] Recently, Cheng et al. 10) have reported that cyanidin inhibited RANKL-induced osteoclast bone resorption via inhibiting RANKL-induced NF-κB, ERK (extracellular signal-regulated kinase) and NFAT activation and calcium oscil-lation, and intraperitoneally injection of cyanidin ameliorated estrogen deficiency-induced bone loss in ovariectomized (OVX) mice, an animal model for postmenopausal osteoporosis. Moriwaki et al. 11) have reported that RANKL-induced osteoclast differentiation was suppressed by delphinidin and cyanidin, and that activity of delphinidin were more potent than that of cyanidin. They also indicated that dietary delphinidin attenuated the bone loss in mouse models of osteoporosis induced by both soluble RANKL (sRANKL) injection and ovariectomy. 11) However, the molecular mechanism and structure-activity relationship of anthocyanidins are still unknown. In this study, we compared the effects of delphinidin, cyanidin and pelargonidin on osteoclast differentiation, and examined the molecular mechanisms of the ACDs in bone resorption.
Organ Cultures of Mouse Calvariae Calvariae of newborn mouse were collected and precultured for 1 d in BGJb medium supplemented with 1 mg/mL of bovine serum albumin (BSA) at 37 °C under 5% CO 2 in air. Calvariae were treated with or without LPS (1 μg/mL) and ACDs (120 μM, each) and cultured for 5 d. The bone-resorbing activity was determined by measuring the calcium concentration in the conditioned medium by o-Cresolphtalein-Complexone (OCPC) methods. The images of absorbed pit areas were scanned by microfocus X-ray CT system (inspXio SMX-90T, Shimadzu Co., Ltd., Kyoto, Japan) and quantified using imaging software (Image J).

Cocultures of POBs and Bone Marrow Cells
Mouse bone marrow cells (BMCs) were collected from tibiae of 6-week-old mice. BMCs were cocultured with POBs with or without LPS (1 μg/mL) and ACDs (120 μM, each) for 7 d, and osteoclasts were stained for tartrate-resistant acid phosphatase (TRAP). TRAP-positive multinucleated osteoclasts were counted as osteoclasts.
Quantitative PCR Analysis To determine the mRNA expression, total RNA was extracted from cells and cDNA was synthesized by reverse transcriptase (Superscript II Preamplification System, Thermo Fisher Scientific Inc., CA, USA) and amplified using quantitative real-time PCR (qPCR). The qPCR analysis (ΔΔCT methods) was performed with iQ SYBR Green Supermix (Bio-Rad Laboratories, Inc., CA, USA). The sequences of primers described in the following; mouse RANKL, (forward) 5'-AGGCTGGGCCAAGATCTC- Beta-actin was used as normalized gene.
IκB Kinase Activity In Vitro Assay The activities of inhibitor of NF-κB kinase (IKK) was measured using IKKα and β Kinase Assay/Inhibitor Screening Kit (Medical & Biological Laboratories Co., Ltd., Nagoya, Japan).
Protein Structure Preparation The three-dimensional X-ray crystal structure of IKKβ was obtained from protein data bank (PDB ID:4KIK, 2.83 Å resolution). 12) For docking simulations, default parameters (H-atoms) were added into the protein structures using AutoDock Tools (Molecular Graphics Laboratory, Torrey Pines Rd. LaJolla, CA, USA).
Ligand Structure Preparation The chemical structure of delphinidin was built and optimized using an online compound editor InDraw (http://in.indraw.integle.com/; Integle Chemistry, Inc., Shanghai, China). All the 2D structures were converted into a three-dimensional structure in the form of pdb format and saved as mol format using Open Babel (http://www.openbabel.org/). 13) Molecular Docking Study The protein-ligand molecular docking study was performed by using AutoDock Vina. 14) Subsequently, the AutoDock Vina was employed to implement fast docking of inhibitor ligand into the active pocket of both IKKβ and kinase domains, which considered flexibility and mobility of the ligand molecules and protein active-site residues, and used Lamarckian genetic algorithm to fully explore conformational space for IKKβ inhibitor interactions. The rotational bonds of the protein were kept rigid,while those of the ligands were treated as flexible. Grid box was set at 60 × 60 × 60 on the x-axis,y-axis, and z-axis. The amino acids Leu21-Tyr 23, Val29, Ala42, Lys44, Glu61, Val74, Met96, Glu97, Tyr98, Cys99, Gly102, Asp103, Glu149, Asn150, Val152, Ile165, Asp166 and the surrounding residues within the distance range of 6.5 Å were defined as active sites.
For the analysisand visualization of protein-ligand inter-action, Pymol (http://apbs.sourceforge.net) was used. Amino acids of active sites and the surrounding residues was selected to show, other parts are hided.

Cultures of Raw264.7 Cells
To examine the differentiation of osteoclast precursor cells into mature osteoclasts, Raw264.7 cells, mouse macrophage cell line, were cultured with or without sRANKL (100 ng/mL) and ACDs (120 μM, each) for 5 d in 96 well plate. To examine the effects of ACDs on the survival of mature osteoclasts, Raw264.7 cells were cultured with sRANKL for 5 d to form mature osteoclasts in 24 well plate, and further cultured with sRANKL and ACDs for 1 d. Osteoclasts were detected by TRAP staining.
Statistical Analysis All data were analysed using one-way ANOVA, followed by Tukey's test for post hoc analysis using IBM SPSS Statistics Ver.23 software (NY, USA).

Effects of ACDs on LPS-induced Osteoclast Differentiation and Bone Resorption
We first examined the effects of delphinidin (Del) on RANKL-induced osteoclast differentiation in Raw264.7 cells, and found that 30-120 μM Del dosedependently suppressed osteoclast formation (supplemental data). Then, we used 120 μM ACDs in the present study. We examined the effects of three ACDs, Del, cyanidin (Cya) and pelargonidin (Pel) (Fig. 1A), on bone resorption in organ cultures of mouse calvariae and osteoclast differentiation in cocultures of POBs and BMCs. To elucidate the bone resorbing activity, the area of bone pit was measured using μCT imaging of calvarial surface. LPS markedly increased the area of pits on calvarial surface, but the pit area was significantly reduced by ACDs (Fig. 1B). In cocultures of BMCs and POBs, Del and Cya significantly suppressed LPS-induced osteoclast differentiation, however, Pel showed no effects (Fig. 1C). We further measured the levels of PGE 2 , an inflammatory mediator, in the conditioned medium of cocultures. ACDs markedly inhibited PGE 2 production induced by LPS (Fig. 1D). These data suggested that ACDs inhibited bone resorption and osteoclast differentiation via the inhibition of PGE 2 production.

Effects of ACDs on PGE 2 Production and RANKL Expression in Osteoblasts
To examine the molecular mechanisms of ACDs in osteoblasts, POBs were cultured with or without LPS and ACDs. PGE 2 production was induced by LPS but ACDs significantly inhibited LPS-induced PGE 2 production ( Fig. 2A). In qPCR analysis, ACDs significantly downregulated LPS-induced mRNA expression of RANKL (Fig. 2B). In reporter gene assay, LPS enhanced NF-κB activiy, however, ACDs blocked its NF-κB activation (Fig. 2C).
Docking Simulation of Delphinidin and ATP Binding Pocket in IKKβ Some natural compounds including flavonoids might bind to ATP-binding pocket of IKKβ to downregulate its kinase activity. 15,16) IKKβ phosphorelated IκBα, an endogenous NFκB inhibitor, and phospho-IκBα are degradated by ubiquitin-proteasome system, leading to nuclear transport of NFκB . In in vitro assay using recombinant IKK without cells, ACDs directly acted on IKK and suppressed its kinase activity (Fig. 2E). In order to get hints about possible binding model of Del, Del was submitted to molecular docking study using a homology model of IKKβ. Fig. 3B showed the overview of the docking simulation workflow. The docking study suggested a blockage caused by Del. As shown in the Fig. 3C, the amino acids Leu21-Tyr 23, Val29, Ala42, Lys44, Glu61, Val74, Met96, Glu97, Tyr98, Cys99, Gly102, Asp103, Glu149, Asn150, Val152, Ile165, Asp166 and the surrounding residues within the distance range of 6.5 Å were defined as active sites and the surrouding area was marked as light blue using pymol software. Del was positioned in catalytic center, a pocket-like structure of the assumed binding site (Fig. 3C).

Effects of ACDs on RANKL-induced Osteoclast Differentiation
We examined the effects of ACDs on osteoclast precursor cells. In Raw264.7 cell cultures, ACDs significantly inhibited sRANKL-induced osteoclast differentiation (Fig. 4A). To test the effects of ACDs on mature osteoclasts, ACDs added to cultures after sRANKL induced osteoclast differentiation in Raw 264.7 cells. ACDs also blocked sRANKLinduced osteoclast survival (Fig. 4B). In qPCR analysis, sRANKL upregulated mRNA expression of NFATc1, CTSK and TRAP, but ACDs attenuated the expression of these genes in Raw 264.7 cells (Fig. 4C).

DISCUSSION
The structure-activity relationship of natural compounds is important to understand their biological activities. In the present study, we newly indicated that Del, Cya and Pel inhibited bone resorption and osteoclast differentiation with a structure-activity relationship. Recent studies have demonstrated the effects of Del and Cya on osteoclast differentiation and bone resorption. Cheng et al. 10) have found that Cya inhibited RANKL-induced osteoclast differentiation via the attenuation of ERK and NFATc1 activation, and further indicated that the intraperitonially injection of cyanidin prevented tibial bone loss in OVX mice. Moriwaki et al. 11) have reported that Del and Cya have been reported to inhibit RANKL-induced osteoclast differentiation in Raw264.7 cells. Hou et al. 8) have reported to compare the effects of five ACDs on cyclooxygenase (COX)-2 expression in macrophages with a structureactivity relationship. They tested five ACDs, Del (3 hydroxyl groups on B ring), Cya (2 hydroxyl groups on B ring), Pel (1 hydroxyl group on B ring), peonidin (1 hydroxyl and 1 methoxy groups on B ring) and malvidin having two methoxy group on B ring. Although Pel, peonidin and malvidin had no effect, Del and Cya significantly reduced LPS-induced COX2 expression in macrophages, and the effect of Del is higher than that of Cya. Since reactive oxygen species (ROS) is crucial factor for osteoclast differentiation, antioxidative activity of ACDs may contribute to the inhibitory effects of osteoclast differentiation. 17) Ali et al. have examined the structure-activity relationships of six ACDs, Del, petunidin, Cya, malvidin, peonidin and Pel, for antioxidative effects. 18) The antiox-  idative activity of these ACDs were depend on the number of hydroxyl and/or methoxy groups, and their strengths of antioxidative activity were described as follows: Del (3 hydroxyl groups on B ring) > petuidin (2 hydroxyl and 1 methoxy groups on B ring) > Cya (2 hydroxyl groups on B ring) > malvidin (1 hydroxyl and 2 methoxy groups on B ring) > peonidin (1 hydroxyl and 1 methoxy groups on B ring) > Pel (1 hydroxyl group on B ring). Our data showed the similar trend: Del showed the most potent activities on inhibition of osteoclast differentiation and bone resorption while that of Pel had lower activities among three ACDs.
In the molecular mechanisms of ACDs, Dou et al. 19) have shown that Cya had the stimulatory effect on RANKL-induced osteoclast differentiation and fusion at the low concentra-tion (< 1 μg/mL) by upregulating the expression of osteoclast marker genes (NFATc1, Mitf and c-Fos) and liver X receptor (LXR)-β, whereas high concentration of Cya (> 10 μg/mL) showed the inhibitory effect by downregulating the expression of osteoclast marker genes. LXR-β was reported to be negative regulator of RANKL-induced osteoclast differentiation, 20) and Cya was shown to be direct agonistic ligand of LXR-β. 21) These reports suggest that LXR-β could be an intracellular target molecule of Cya to inhibit osteoclast differentiation. Our present study indicated IKKβ was the novel target molecule of ACDs, and ACDs directly attenuated IKKβ activities with a structure-activity relationship. Kim et al. 15) have suggested that luteolin, one of flavonoid, might bind to ATP-binding pocket of IKKβ to downregulate its kinase activity. Our dock- ing simulation of Del and IKKβ indicated that ACDs could bind to ATP-binding pocket of IKKβ to suppress its kinase activity. Considering the similarity of the structure of Del and K252a, a known ATP-competitive IKKβ inhibitor, 12) Del might contact with the hinge region of IKKβ by forming a hydrogen bond between the Del and IKKβ. Since IKK is required for both signalling of LPS-TLR4-NF-κB and RANKL-RANK-NF-κB, ACDs may block IKK-dependent NF-κB activation in both osteoblasts and osteoclasts, leading to the suppression of osteoclastic bone resorption. Since bone turnover is regulated by osteoblastic bone formation and osteoclastic bone resorption, it is possible that ACDs may act on osteoblasts to regulate bone formation. We will plan to examine the effects of ACDs on osteoblast differentiation in the near future.
In conclusion, we have shown that the activities of ACDs on osteoclastic bone resorption is dependent on the total number of hydroxyl group on B ring, and IKK may be the novel intracellular target molecule of ACDs. Del may be the most beneficial compounds for preventing of bone destructive diseases in three major ACDs, Del, Cya and Pel.