Chondroitin sulfate-E mediates estrogen-induced osteoanabolism

Osteoporosis is an age-related disorder of bone remodeling in which bone resorption outstrips bone matrix deposition. Although anticatabolic agents are frequently used as first-line therapies for osteoporosis, alternative anabolic strategies that can enhance anabolic, osteogenic potential are actively sought. Sex steroid hormones, particularly estrogens, are bidirectional regulators for bone homeostasis; therefore, estrogen-mediated events are important potential targets for such anabolic therapies. Here, we show that estrogen-induced, osteoanabolic effects were mediated via enhanced production of chondroitin sulfate-E (CS-E), which could act as an osteogenic stimulant in our cell-based system. Conversely, estrogen deficiency caused reduced expression of CS-E-synthesizing enzymes, including GalNAc4S-6ST, and led to decreased CS-E production in cultures of bone marrow cells derived from ovariectomized mice. Moreover, Galnac4s6st-deficient mice had abnormally low bone mass that resulted from impaired osteoblast differentiation. These results indicated that strategies aimed at boosting CS-E biosynthesis are promising alternative therapies for osteoporosis.

estrogen-mediated cellular events, including osteoblastic bone formation. Here, we aimed to test whether osteoanabolic activities of estrogens are exerted via CS-E production.

Results
Estrogen induces CS-E production in osteoblastic cells. To test our hypothesis, we first used mouse MC3T3-E1 cells to represent osteoblastic cells 31 . MC3T3-E1 cells express N-cadherin and cadherin-11, and cadherin-mediated cell-cell contact is critical for the onset of osteoblastic differentiation of these cells 32,33 . Once committed to differentiate, MC3T3-E1 cells produce increasing amounts of CS with a relatively high proportion of E units 24 . Therefore, we treated low-density MC3T3-E1 cultures with estradiol, a predominant form of estrogen, to avoid undesirable CS production that is induced by cell-cell contact. After a 24-h exposure to estradiol, when individual cells remained dispersed, CS isolated from MC3T3-E1 cultures was depolymerized with a bacterial CS-degrading enzyme, chondroitinase ABC (ChABC), and resultant CS disaccharides were analyzed via high-performance liquid chromatography (Table 1). Treatment with estradiol resulted in approximately a 20% increase in the total content of CS disaccharides in low-density cultures; this increase was mainly attributable to elevated levels of 4-O-sulfated CS disaccharides, A [GlcA-GalNAc(4-O-sulfate)] and E units. In contrast, the amount of 6-O-sulfated disaccharide C [GlcA-GalNAc(6-O-sulfate)] units was essentially unaltered. This trend was quite similar to that obtained in MC3T3-E1 cultures undergoing differentiation induced by cell-cell contact 24 .
These sulfated CS disaccharides (A, E, and C units) are formed during CS biosynthesis 12 . Briefly, precursor disaccharide O units [GlcA-GalNAc] in the chondroitin backbone can serve as common acceptor substrates for two types of sulfotransferases (CHSTs), chondroitin 4-O-sulfotransferases (C4STs) or chondroitin 6-O-sulfotransferase-1 (C6ST-1). C4STs catalyze 4-O-sulfation and C6ST-1 catalyzes 6-O-sulfation of GalNAc residues to form A or C units, respectively; subsequent sulfation of A units by an additional enzyme, GalNAc4-sulfate 6-O-sulfotransferase (GalNAc4S-6ST), generates E units. Therefore, sulfations of a chondroitin backbone can be classified into ''4-O-sulfation'' or ''6-O-sulfation'' pathways ( Fig. 1a). To examine whether the apparently exclusive increase in 4-O-sulfated CS containing E units in the estrogen-treated cultures resulted from alteration of CS biosynthetic machinery, expression levels of genes encoding the abovementioned CHSTs were measured. Consistent with the biochemical data for CS, expression of two genes, C4st1 and Galnac4s6st, that encode enzymes involved in 4-Osulfation pathway was significantly upregulated, whereas gene expression relevant to 6-O-sulfation was not affected by estradiol treatment (Fig. 1b). Additionally, expression of transcripts encoding ChGn-1 and ChGn-2 did not change significantly (Fig. 1a,c); these enzymes, along with C4STs, catalyze synthesis of the chondroitin backbone and control CS levels 34,35 . These findings indicated that the estradiol-induced increases in CS were ascribable to augmentation of C4st1 expression. Notably, basal expression of Akp2, a typical osteogenic marker gene encoding alkaline phosphatase (ALP), remained constant in the 24-h cultures, even in the presence of estradiol (Fig. 1d). Thus, the dynamic changes in expression of genes encoding CS biosynthetic enzymes were not due to secondary effects associated with accelerated differentiation of MC3T3-E1 cells, but instead to possible direct regulation by estradiol.
To examine the generality of the observed phenomena, we examined bone marrow-derived stromal cells (BMSCs) instead of MC3T3-E1 cells, because BMSCs have a significant population of osteoblast progenitors that can contribute to bone formation in vivo. In low-density cultures of BMSCs, bath-applied estradiol resulted in a similar pattern of changes in expression of mRNA encoding CHSTs ( Supplementary Fig. S1). Thus, these results suggested that estrogen preferentially induces the expression of distinct CS subtypes that have a relatively high amount of E units (hereafter CS-Es) in osteoblastic cells.
Reduced estrogen level causes reduced CS-E production. To evaluate in vivo effects of estrogens on CS-E production in boneanabolic milieus, female mice were ovariextomized (OVX) to deplete endogenous estrogens. At 6 weeks after ovariectomy, BMSCs were isolated from femora of either OVX or sham-operated mice, and CS production in the isolated BMSCs was measured. Compared to cells isolated from sham-operated controls (Sham), BMSCs derived from OVX mice exhibited approximately a 50% reduction in CS level and a prominent decrease in the amount of E units ( Table 2). Consistent with this observation, levels of C4st1 and Galnac4s6st, genes encoding E-unit-synthesizing enzymes, were significantly lower in BMSCs from OVX mice than those from Sham mice (Fig. 2a). In contrast, ovariectomy did not affect expression of Csgalnact1 or Csgalnact2 (Fig. 2b); these findings were reminiscent of those from estradiol-treated MC3T3-E1 cultures (Fig. 1b,c and Supplementary  Fig. S1). Additionally, microcomputed tomography (mCT) analysis of tibias, which were excised from respective mice in the same time window as BMSC isolation, was performed to confirm that OVX mice exhibited characteristics of mouse-modeled postmenopausal osteoporosis. Indeed, OVX mice had significantly lower bone mineral density (BMD) in both cortical and trabecular bones than did control mice (Fig. 2c). These findings indicated that CS-E production in bone-forming cells was strictly regulated by estrogens in vivo.
Mice deficient in GalNAc4S-6ST have low bone mass. Based on our notion that CS-E production is a critical event for estrogen-mediated bone anabolism, we expected that genetic deficiencies of C4ST-1 and GalNAc4S-6ST, two enzymes involved in the biosynthesis of E units, would have osteopenic/osteoporotic phenotypes. A gene-trap mutation in C4st1 causes severe skeletal phenotypes in mice 22 ; nevertheless, direct effects of this mutation on bone formation might be difficult to assess because the mutation causes marked chondrodysplasia phenotypes and neonatal lethality. Interestingly, mice lacking GalNAc4S-6ST (Galnac4s6st 2/2 ) that result in a complete loss of CS-E disaccharides show no readily apparent developmental abnormalities 36 , but detailed phenotypic analyses of Galnac4s6st 2/2 mouse skeletons are not published. Therefore, we carefully documented detailed skeletal phenotypes of Galnac4s6st 2/2 mice.
At 16 weeks after birth, male Galnac4s6st 2/2 mice had significantly lower BMD than male wild-type (WT, Galnac4s6st 1/1 ) littermates; a significant difference between Galnac4s6st 1/2 and Galnac4s6st 1/1 littermates was also observed in the total BMD (Fig. 3a). Uniform  decrease in BMD along the tibia of Galnac4s6st 2/2 mice indicated that both cortical and trabecular bones were affected equally ( Supplementary Fig. S2). Using mCT three-dimensional reconstruction of tibias, we confirmed these observations (Fig. 3b). Thus, impaired synthesis of CS-E caused both cortical and trabecular bone loss in adult (at least 16-week-old), male mice.
Regulatory roles of CS-E in bone remodeling. To examine whether the osteopenic/osteoporotic phenotypes of Galnac4s6st 2/2 mice were caused by defects in bone anabolism, the osteoblastic potential of BMSCs, derived either from WT or Galnac4s6st 2/2 mice, were assessed. As reported previously 36 , we confirmed that Galnac4s6st 2/2 BMSCs also produced CS completely devoid of E units (Table 3). Osteoblast differentiation and maturation constitute a highly ordered process that begins with ALP expression and ends with mineral deposition; therefore, isolated BMSCs were cultured in an osteogenic medium; on days 4 and 21, cells were stained for ALP and mineralized nodule formation, respectively. ALP expression was significantly lower in Galnac4s6st 2/2 BMSC cultures than in WT controls (Fig. 4a,c). Mineralized nodule formation in Galnac4s6st 2/2 BMSC cultures was also severely impaired (Fig. 4d). Notably, bath application of estradiol led to a significant increase in ALP expression and mineral deposition in WT BMSC cultures, but not in Galnac4s6st 2/2 BMSC cultures ( Fig. 4a-d). We previously showed that exogenous addition of CS-E polysaccharides could increase ALP expression in the low-density MC3T3-E1 cultures 24 , and such a stimulatory effect was not compensated by heparin, another class of highly sulfated GAGs ( Supplementary Fig. S3). Consistent with these findings, bath-applied CS-E, but not heparin, significantly augmented ALP expression even in initial 24-h cultures of Galnac4s6st 2/2 BMSCs (Fig. 4f), although the ALP level remained lower than that obtained from intact Galnac4s6st 1/1 BMSCs cultures (Fig. 4e,f). These data indicated that CS-E formed by GalNAc4S-6ST is essential for osteoblast differentiation and maturation, and that estrogens are potent inducers of CS-E-mediated osteoblastgenesis.
We also evaluated the osteoclastic potential of bone marrowderived macrophages (BMMs) isolated from femoral and tibial long bones of WT or Galnac4s6st 2/2 mice. In the presence of RANKL (receptor-activator of NF-kB ligand), an osteoclast differentiation factor, BMMs can differentiate into tartrate-resistant acid phosphatase (TRAP)-positive, multinucleated osteoclasts in vitro. Based on indexes of osteoclast fusion, we found no significant differences between WT and Galnac4s6st 2/2 BMM cultures ( Supplementary  Fig. S4). This result excluded the possibility that increased osteoclastgenesis in Galnac4s6st 2/2 mice might lead to the osteopenic phenotypes, and strengthens the hypotheses that CS-E contributes substantially to bone formation, but not to bone resorption.

Discussion
Estrogens exhibit pleiotrophic effects on bone homeostasis including bone formation and resorption 9,10 . In the present study, we investigated distinctive downstream events of estrogen-mediated osteoanabolism. Our findings demonstrate that CS-E production was critical to proper bone anabolism and that this CS-E production was regulated by estrogens. Galnac4s6st 2/2 mice exhibited an undetectable CS-E expression in the bone marrow niche, and experienced significant bone loss because of impaired osteoblastic potential. These findings led to the novel idea that postmenopausal osteopenia and osteoporosis are caused by both excessive osteoclast formation and reduced anabolic bone formation, and that this anabolic activity was mediated by estrogen-induced CS-E. Our previous findings demonstrate that CS-E promotes osteoblast differentiation of MC3T3-E1 cells via its binding to cell surface N-cadherin and cadherin-11 24 . Additionally, osteoblast-and osteocyte-specific conditional knockout of Cdh2 (N-cadherin) and double Cdh2 1/2 ; Cdh11 2/2 mutation reportedly lead to osteopenia in adult mice 37 . Hence, estrogeninduced CS-E may also exert anabolic effects on living bone tissue via distinct cadherin-mediated signaling pathways. Since CS production is also up-regulated by cadherin-mediated cell-cell contact 24 , such an initial stimulation of cadherin pathways by estrogen-induced CS-E may induce subsequent CS production to form optimal osteoanabolic milieus. Notably, bath-applied CS-E can activate intracellular signaling required for osteogenesis of MC3T3-E1 cells even in low-density cultures, where cadherin-dependent cell-cell contact does not occur; moreover, these CS-E effects are abrogated by blocking the respective cadherins 24 . Likewise, in Galnac4s6st 2/2 BMSCs, exogenous CS-E could also stimulate initial onset of osteoblast differentiation. These findings suggest a mode of action for CS-E; specifically, CS-E may act as functional ligands for two cell surface CS receptors, N-cadherin and cadherin-11. Therefore, CS-E may be useful as pharmacological agents that promote osteogenesis in patients suffering from metabolic bone diseases, including osteopenia and osteoporosis. In addition to CS-E, CS-A level was also up-regulated in estrogen-treated osteoblastic cultures. In contrast, it was dramatically decreased in BMSCs obtained from OVX and   Estrogen replacement therapy is a widely accepted treatment for osteoporosis 8 . However, estrogens have multiple sites of action; consequently, hormonal therapy can increase the risk of thromboembolic events and breast and uterine cancers; accordingly, long-term use of estrogen replacement therapy is limited 38 . To overcome this problem, tissue-targeted hormonal replacement has been proposed. Additionally, strict control of events downstream of estrogens (tissue-specific, function-specific, or both) may become effective therapeutic approaches. In this context, regulation of CS-E expression in bone metabolism may have substantial therapeutic potential. Nevertheless, how CS-E production is controlled by estrogens remains unknown, and direct or indirect transcriptional activation of the E-unit synthesizing enzymes via estrogen receptors is expected to be involved in the process. Furthermore, since CS chains are expressed as CSPGs, identification of core proteins carrying CS-E in the bone marrow niche is also an essential task. Interestingly, double knockout of Bgn (biglycan) and Dcn (decorin), members of class I type small leucine-rich PGs, results in reduced bone formation and shows osteoporosis-like phenotypes 39 . Therefore, the underlying mechanisms for functional expression of E-unit synthesizing enzymes and the core protein candidates, and the cell and/or tissue specificity of such mechanisms must be clarified for development of CS-E-based therapies. In conclusion, therapies that enhance CS-E production could become highly effective treatments that ameliorate   DDi-0S 40.   Eight-week-old virgin female WT mice were subject to either bilateral sham operation or ovariectomy and then reared for 6 weeks before use.
Osteoblast cultures. Mouse osteoblastic MC3T3-E1 cells (RCB1126) were obtained from RIKEN Cell Bank (Tsukuba, Japan), and these at passages 3-10 were used in this study. BMSCs were harvested from long bones (femora plus tibias or femora alone) of 14-week-old WT female mice from the intact, sham-operated, or ovariectomized group and from femora of 16-week-old WT or Galnac4s6st 2/2 male mice as described 40 . Based on the representative flow cytometric profiles ( Supplementary Fig.  S5), the isolated BMSCs were largely free of contamination of endothelial and hematopoietic cells. Mouse osteoblastic MC3T3-E1 cells or BMSCs from WT intact female mice were seeded at a low density (5 3 10 3 cells cm 22 ) and incubated for 24 h in growth medium (GM, aMEM containing 10% FBS) in the presence or absence of 10 nM estradiol. To assess the influence of estrogen deficiency, BMSCs from shamoperated or ovariectomized WT female mice were cultured in GM for 3 days. BMSCs from WT or Galnac4s6st 2/2 male mice were placed on 96 well plates at a high density (1 3 10 4 cells per well) and cultured for 21 days in differentiation medium (DM, GM supplemented with 100 mg ml 21 ascorbic acid, 10 mM b-glycerophosphate and 10 nM dexamethasone) to elicit osteoblast differentiation 24 . In some cases, BMSCs were maintained in DM supplemented with 10 nM estradiol for 4 days. ALP staining 24 for cellular ALP activity and alizarin red staining for measurement of mineral matrix formation were performed at days 4 and 21, respectively. To examine the effects of exogenous GAGs on osteoblast differentiation, low-density cultures of MC3T3-E1 cells or Galnac4s6st 2/2 BMSCs were performed for 24 h in GM in the presence or absence of GAGs (CS-E or heparin, 20 mg/ml each).
Osteoclast cultures. Bone marrow cells were harvested from femora and tibias of 8 to 12-week old WT or Galnac4s6st 2/2 male mice. BMMs were concentrated from bone marrow cells as described 41 . To induce osteoclast differentiation, the resultant BMMs (2 3 10 4 cell per well of 96-well plates) were stimulated with M-CSF (50 ng ml 21 ), TGF-b (1 ng ml 21 ), and RANKL (300 ng ml 21 ) for 72 h. To evaluate osteoclast formation, cells were stained for TRAP 41 .
Disaccharide composition of CS. CS from MC3T3-E1 and BMSC cultures was prepared as described previously 42 . The purified glycosaminoglycan fraction containing CS was digested with chondroitinase ABC at 37uC for 2 h. The digests were derivatized with the fluorophore 2-aminobenzamide and were then analyzed via anion-exchange HPLC on a PA-03 column (YMC, Kyoto, Japan). Identification and quantification of the resulting disaccharides were achieved by comparison with authentic unsaturated CS disaccharides (Seikagaku).
Statistical analysis. Unless otherwise specified, statistical significance was determined via two-tailed Student's t test. Differences were considered to be significant with a P value less than 0.05.
Ethics statement. All animal procedures were approved by the Kobe Pharmaceutical University Committee on Animal Research and Ethics. All experiments were conducted in accordance with the institutional ethical guidelines for animal experiments and safety guidelines for gene manipulation experiments.