3-Ketosteroid Reductase Activity and Expression by Fetal Rat Osteoblasts*

In addition to reproductive tissue, sex hormones induce transcriptional events in many connective tissue cells, including osteoblasts. Some sex hormone receptor modulators with bone sparing effects selectively target estrogen or androgen receptors, whereas others appear more promiscuous, in part through enzymatic metabolism. Rat osteoblasts express significant oxidative 3α-hydroxysteroid dehydrogenase activity, which can convert precursor substrates to potent androgen receptor agonists. Here we show that they also express 3-ketosteroid reductase activity, exemplified by 7-methyl-17-ethynyl-19-norandrostan-5 (10)en-3-one (tibolone) conversion to potent estrogen receptor α agonists. Conversion was rapid and quantitative, with 3α-hydroxytibolone as the primary metabolite. Consistently, tibolone induced estrogen receptor α-dependent gene promoter activity through cis-acting estrogen response elements, increased the stimulatory effect of TGF-β on Smad-dependent gene promoter activity, and enhanced prostaglandin E2-induced activity of transcription factor Runx2. Rat osteoblasts express the 3-ketosteroid reductase AKR1C9, an aldo-keto reductase gene family member. Exposure to prostaglandin E2 increased AKR1C9 gene promoter activity and mRNA expression. AKR1C9 promoter activity was also enhanced by overexpression of protein kinase A catalytic subunit or transcription factor C/EBPδ, and the effect of PGE2 was reduced by dominant negative C/EBPδ competition or C/EBPδ antisense expression. Moreover, prostaglandin E2 increased the amount of functional endogenous nuclear C/EBPδ that could bind specifically to a distinct domain ∼1.8-kb upstream from the start site of AKR1C9 transcription. In summary, in addition to 3α-hydroxysteroid dehydrogenase, rat osteoblasts express significant and regulatable 3-ketosteroid reductase activity. Through these enzymes, they may selectively metabolize precursor compounds into potent steroid receptor agonists locally within bone.

possible regulatory cis-acting elements. To date, only few of these elements have been addressed, and essentially all that is known derives from studies in liver cells where the enzyme was initially thought to have its primary activity (13)(14)(15). Tibolone (7-methyl-17-ethynyl-19-norandrostan-5 (10)en-3-one), a 3-ketosteroid androgen receptor (AR) agonist with potential HRT efficacy, has complex effects in vivo through AR, ER, and progesterone receptor (PR), in part through rapid conversion to other metabolites in the organism. Reduction of tibolone at position 3 in steroid ring A by AKR1C family members produces the ER agonists 3␣-hydroxytibolone and 3␤-hydroxytibolone (16 -18). In this study we assessed the activation potential of tibolone on ER␣-dependent gene induction in rat osteoblasts, using it as a model for endogenous 3-ketosteroid reductase activity in bone. Based on previously unrecognized hormone interactions and on sequence analyses, we also assessed AKR1C9 expression and gene promoter activity in these cells to define molecular mechanisms that drive its expression. Our results show that osteoblasts, through inherent 3-ketosteroid reductase activity, can metabolize steroid substrates into potent estrogens or possibly to limit their androgenic activity. They further predict that hormone-dependent changes in 3-ketosteroid reductase expression by osteoblasts could oppose their endogenous 3␣-/3␤-hydroxysteroid dehydrogenase oxidation potential, ultimately regulating the levels of ER or AR agonists within the skeletal tissue environment.

EXPERIMENTAL PROCEDURES
Cells-Primary osteoblast-enriched cultures were isolated from parietal bones of 22-day-old Sprague-Dawley rat fetuses (Charles River Breeding Laboratories), as approved by the Yale Institutional Animal Care and Use Committee. Sutures were dissected, and cells were released by five sequential collagenase digestions. Cells pooled from the last three digestions express features of differentiating osteoblasts, including high levels of runt homology domain nuclear factor 2 (Runx2), parathyroid hormone receptor, type I collagen synthesis, and alkaline phosphatase. They also increase osteocalcin expression in response to vitamin D 3 , exhibit differential sensitivity to transforming growth factor ␤ (TGF-␤), bone morphogenetic protein 2, and various prostaglandins (PGs), and form mineralized nodules under conditions promoting long term differentiation in vitro. Cells were plated at 4,000/cm 2 in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 100 g/ml ascorbic acid and grown for 6 days before transfection or treatment, which was performed in serum-free medium (19 -26).
Transfection Plasmids-AR-dependent gene expression was assessed with a firefly luciferase reporter plasmid driven by four consensus androgen response elements (ARE) cloned upstream of a minimal RSV promoter, and ER␣-dependent gene expression was assessed with a luciferase reporter plasmid driven by an estrogen response element (ERE) from the frog vitellogenin promoter cloned upstream of a minimal prolactin gene promoter in cells co-transfected with expression plasmids encoding AR or ER␣ (12). PR-dependent activity was assessed with a luciferase reporter plasmid driven by two progesterone response elements cloned upstream of a minimal TATA box in cells cotransfected with PR-A or PR-B (27,28). Runx activity was assessed with luciferase reporter plasmid 5XGAL4 driven by five GAL4 response elements in cells co-transfected with an expression plasmid encoding a Runx2-GAL4 DNA binding domain fusion protein (M1Runx2) (12,29). Smad-dependent gene expression was assessed with luciferase reporter plasmid SBE4 driven by four Smad response elements (29). AKR1C9 gene promoter activity was assessed with a 2.0-kb fragment of the rat promoter (GenBank TM AF180326) (13) subcloned into luciferase reporter plasmid pGL2-Basic. CCAAT enhancerbinding protein (C/EBP) ␦ and ␤ overexpression were induced with expression plasmids encoding the rat mRNAs. Endogenous C/EBP activity was suppressed by transfection with an expression plasmid encoding a dominant negative rat C/EBP␦ (dn-C/EBP␦) fragment encompassing its leucine zipper dimerization and DNA-binding domains but no gene transactivation domain (30,31), or rat C/EBP␦ in the reversed, antisense orientation (␣s-C/EBP␦). To assess the role of PKA, cells were co-transfected with expression plasmids encoding catalytic, regulatory, or dominant negative subunits of PKA (32)(33)(34)(35)(36)(37).
Transfections-Promoter-reporter fusion plasmids, gene expression plasmids, or empty parental vectors, were pretitrated for optimal expression efficiency and transfected with reagent LT1 (Mirus). Cultures at 70% confluence were exposed to an optimal amount of reporter plasmid (75 ng per cm 2 ) or expression plasmid (10 -75 ng per cm 2 ) in medium supplemented with 4% serum. Cells were cultured for 6 -72 h and treated in serum-free medium as indicated in the figures. The cells were rinsed and lysed, and supernatants were analyzed for reporter gene activity and corrected for protein content. To account for competition among plasmids for limiting transcriptional components, control cultures were transfected with a compensating amount of empty vector. Transfection efficiency was assessed in parallel in cells transfected with positive and negative reporter plasmids as previously described (23,38).
HPLC-To assess tibolone metabolites, an elution profile for purified 3 H-labeled tibolone and its known conversion products (17,18)  Bondapak. Samples were applied in 60% methanol and eluted at 1 ml/min for 5 min, followed by a 20-min linear gradient to achieve 65% methanol, and continued at 65% methanol for 15 additional minutes. Elution was monitored by liquid scintillation counting. A combined elution profile of the four readily separated compounds is shown in the upper panel of Fig. 2A, where tibolone migrates at 25 min, ⌬ 4 -tibolone at 17 min, 3␣-hydroxytibolone at 20 min, and 3␤-hydroxytibolone at 28 min. Osteoblasts were then incubated with the 0.5 Ci of [ 3 H]tibolone or its derivatives for 0, 2, 6, or 24 h. Each culture medium was combined with a methanol extract from its cell layer and extracted with hexane. The aqueous fraction was ether extracted and steroids in the ether extract were analyzed by HPLC as described above. Extraction efficiency and recovery was determined with reference to 2 g of unlabeled ⌬ 4 -tibolone as an internal UV light (240 nM) absorbent recovery standard.
All tibolone and tibolone derivatives used in these studies were generously supplied by Organon NV, Oss, The Netherlands.
mRNA Analysis-Total RNA was extracted with acid-guanidine-monothiocyanate, precipitated with isopropyl alcohol, and dissolved in sterile water. AKR1C9 mRNA was assessed by fractionation on a 1.5% agarose/2.2 M formaldehyde gel, blotting on charged nylon, and hybridization with 32 P-labeled cDNA encompassing a 0.6-kb HindIII/EcoRI restriction fragment (bp 251-854) of the coding region of rat AKR1C9 (12). rRNA was assessed by ethidium staining of a parallel gel. Radiolabeled products were examined by autoradiography and densitometry (39).
Electrophoretic Mobility Shift Analysis (EMSA)-Doublestrand oligonucleotide probes comprising AKR1C9 promoter fragments that define three separate C/EBP binding sites ( Table  1) were labeled with [ 32 P]dCTP and Klenow fragment of Escherichia coli DNA polymerase I, and gel-purified. Nuclear protein extracts (3 g) from control or PGE2 induced cells were preincubated with either no addition, a 100-fold excess of unlabeled oligonucleotide (Table 1), nonimmune IgG, or antiserum to C/EBP␦ or C/EBP␤ (Santa Cruz Biotechnology, Inc.), and then supplemented with 32 P-labeled probe. Protein-bound DNA complexes were resolved on a 5% nondenaturing polyacrylamide gel and examined by autoradiography (39,40).
Statistical Analysis-Differences were assessed by one-way analysis of variance with Tukey post hoc analysis using Sigma-Stat (Jandel Corporation) from nine or more replicate samples and two or more independent cell preparations. Results from mRNA and gel shifts analyses were from at least two studies. A significant difference was assumed by a p value of Ͻ0.05.

Sex Steroid Receptor-dependent Gene Expression by Tibolone
in Osteoblasts-SSR agonists induce direct and indirect effects on gene expression in osteoblasts in relatively receptor restricted ways (12,29). Unlike the focused effects of dihydrotestosterone (DHT) or 17␤-estradiol (17␤E), exposure to tibolone produced direct, dose-dependent effects through both AR and ER␣ in osteoblasts. The effect of tibolone closely paralleled the AR-specific response to DHT through ARE (Fig. 1A, upper panel) and the ER␣-specific response to 17␤E through ERE (Fig. 1A, lower panel). In addition, like 17␤E (12,29), tibolone enhanced the activity of the essential osteoblast transcription factor Runx2 in combination with ER␣ but not AR (Fig.  1B). These results revealed that tibolone, itself an AR selective agonist (16), achieved ER␣ agonist potential in the presence of metabolically active osteoblasts.
Osteoblasts -tibolone (data not shown). Thus, through conversion or inter-conversion, 3␣-hydroxytibolone accounted for the primary tibolone metabolite, albeit some was lost to other polar, but presently unknown breakdown products. Moreover, these results establish with certainty that osteoblasts express significant 3-ketosteroid reductase activity in addition to their oxidative 3␣/␤hydroxysteroid dehydrogenase activity (12).

Osteoblast 3-Ketoreductase Activity and Expression
in osteoblasts resulting from its rapid reduction to 3-hydroxytibolone metabolites. When osteoblasts were treated with tibolone in the presence of two different oxidoreductase inhibitors, quercetin or phenolphthalein (41,42), its ER␣ activation potential was severely reduced. For presently unexplained reasons, each inhibitor enhanced basal ER␣-dependent gene expression in the absence of ligand, but the significant stimulatory effect of tibolone was nonetheless reduced by ϳ70% (Fig. 2D).
Consistent with in vitro receptor binding (16), tibolone and ⌬ 4 -tibolone also potently enhanced progesterone response element-driven gene expression through PR-B in osteoblasts, 3␤-hydroxytibolone had more modest activity, and 3␣-hydroxytibolone had no significant effect. By contrast to results with PR-B, highly attenuated effects occurred through PR-A (supplemental Fig. S1). This pattern of agonist-dependent activity through PR-B was analogous to their AR-dependent effects through ARE in osteoblasts, and to their biochemical effects in human endometrial cancer-derived cells (43).
PGE2 Enhances Tibolone-dependent Runx2 Activity through ER␣-PGE2 differentially regulates osteoblast activity in complex ways (21,34,44,45). This can occur in part through an increase in Runx2 activity (32), which may be augmented further in complex with hormone-activated ER␣ (29). Tibolone also superenhanced Runx2 activity in PGE2-activated osteoblasts analogous to the effects of 17␤E or the 3␣/␤-hydroxytibolones, as predicted from its ER␣-activating potential after 3-ketosteroid reductase-dependent reduction (Fig. 3, left panel). There was no significant increase in PGE2 activity by any AR agonist including tibolone (Fig. 3, right panel) or its metabolites (data not shown).
PGE2 Increases AKR1C9 Gene Expression in Osteoblasts-Sequence analysis with MatInspector (Genomatix Software, GmbH), revealed that the 2.0-kb 3Ј-region of the AKR1C9 gene promoter contains several possible domains associated with PGE2-induced transcription factors. Accordingly, PGE2 potently enhanced AKR1C9 gene promoter activity in osteoblasts. The stimulatory effect of PGE2 was dose-and time-related, achieving an approximate 4-fold increase within 6 h with 1 M PGE2 (Fig. 4A, left panel), and declined significantly by 24 h (Fig. 4A, right panel). The effect of PGE2 was mimicked by forskolin but not by the phorbol ester PMA (Fig. 4B), predicting gene activation by a protein kinase A (PKA)-sensitive pathway. Osteoblasts expressed a single species of AKR1C9 mRNA of ϳ2.7 kb, consistent with the size calculated by genomic analysis (46). In agreement with gene promoter activation, densitometric analysis indicated that AKR1C9 mRNA levels increased by 3.4 Ϯ 0.1-fold within 6 to 12 h, and decreased to 1.7-fold after 15 h of PGE2 treatment (Fig. 4C).
Overexpression of the catalytic subunit of PKA also increased AKR1C9 gene promoter activity in a time-dependent FIGURE 2. Tibolone is rapidly reduced to 3-hydroxy derivatives in osteoblasts. In A, purified samples of [ 3 H]tibolone and its derivatives were chromatographed by HPLC using a methanol gradient as shown in the upper panel, generating the standard elution profile compiled in the panel directly below. Osteoblasts were then incubated with [ 3 H]tibolone, and the resulting compounds found in the cultures were reanalyzed after 0, 2, and 6 h, as indicated in the lower three panels. In B, osteoblasts were co-transfected for 24 h to express AR and reporter plasmid ARE, or ER␣ and reporter plasmid ERE as in Fig. 1, and then treated for 24 h with vehicle (0) or 10 nM tibolone (Tib), 3␣-hydroxytibolone (3␣-Tib), 3␤-hydroxytibolone (3␤-Tib), ⌬ 4 -tibolone (⌬4-Tib), DHT, or 17␤E as indicated. In C, the cells were co-transfected to express ER␣ and reporter plasmid SBE4, and treated for 24 h with vehicle (0) or 10 nM Tib, and then supplemented with vehicle (Ϫ) or 0.12 nM TGF-␤ (ϩ) for another 24 h. In D, the cells were co-transfected to express ER␣ and reporter plasmid ERE, and treated with vehicle (Ϫ) or 1 nM Tib (Ϫ) for 24 h in combination with the AKR inhibitors quercetin (QCT) at 0.1 mM or phenolphthalein (PPHT) at 10 M, as indicated. Reporter activity was measured after that last 24-h treatment interval. The numbers in parentheses above the filled bars in C and D indicate the fold stimulatory effect of tibolone or 17␤E relative to TGF-␤ activity in the absence of steroid. The numbers above the filled bars in D indicate the fold stimulatory effect of tibolone without (control) or with QCT and PPHT. All compounds except 3␣-hydroxytibolone and 17␤E significantly enhanced gene expression through ARE in AR-transfected cells, and all compounds except ⌬ 4 -tibolone and DHT significantly enhanced gene expression through ERE in ER␣-transfected cells. Tibolone significantly enhanced TGF-␤-induced SBE4 activity, and the stimulatory effect of tibolone through ERE was significantly reduced by both AKR inhibitors in ER␣-transfected cells (p Ͻ 0.05). way. Although basal AKR1C9 promoter activity increased during 72 h of culture, it was significantly enhanced during the first 2 days of expression by co-expression of the catalytic subunit of PKA. Like the biphasic effect of PGE2, the stimulatory effect of the PKA catalytic subunit regressed to base line levels after continued expression, suggesting induction by way of a tran-siently activated transcriptional component (Fig. 5A, left panel). Moreover, the stimulatory effect of PGE2 was blocked by 80% by overexpression of a dominant negativemutated PKA regulatory subunit that cannot bind cAMP and release active catalytic subunit (Fig. 5A,  right panel). Like PGE2, transgenic expression of the PKA catalytic subunit enhanced AKR1C9 mRNA expression, and expression of the mutated regulatory subunit of PKA blocked the stimulatory effect of PGE2 (Fig. 5B). Thus, PGE2 appears to increase AKR1C9 gene expression in large part through an endogenous PKA-sensitive transcriptional component.

C/EBP␦ Regulates the AKR1C9 Gene Promoter in Osteoblasts-
Several PKA-sensitive transcription factors occur in fetal rat osteoblasts, including CREB (47), the C/EBPs (31,48), Runx2 (32,49), and Fra2 (33). Of these, CREB, C/EBP, and Runx2 binding sites occur within the 3Ј 2.0-kb AKR1C9 promoter region. CREB and Runx2 are constitutively expressed at high levels in differentiated osteoblasts, but within several contexts their abilities to bind DNA do not appear to be PGE2-sensitive (32,47). However, Runx2 drives the expression of C/EBP␦ (32), which then accumulates in the nucleus in osteoblasts and induces gene expression after exposure to PKA activating hormones like PGE2 (31,50). The possibility for involvement by C/EBP␦ was therefore assessed within the context of the 2.0 kb AKR1C9 promoter region. Transgenic overexpression of C/EBP␦ significantly enhanced AKR1C9 gene promoter activity, and the stimulatory effect of PGE2 was also greater in C/EBP␦-overexpressing cells (Fig. 6A, left panel). Furthermore, overexpression of a dominant negative C/EBP␦ fragment encoding leucine zipper dimerization and DNA binding domains but no gene transactivation domain (31,39) significantly limited the effect of endogenous C/EBP␦ on AKR1C9 gene promoter activity and reduced the effect of PGE2 by 80 -90% (Fig. 6A, middle panel). Co-expression with a vector containing C/EBP␦ in antisense orientation to limit C/EBP␦ mRNA levels also significantly limited AKR1C9 gene promoter activity in PGE2 induced cells, by ϳ75% (Fig. 6A, right panel). This was only slightly less effective than the inhibitory effect of the dominant negative C/EBP␦ construct, which directly targets and suppresses stimulation by the pre-existing pool of functional C/EBP␦ protein that occurs in differentiating osteoblasts (31,32,47), perhaps due to differences between C/EBP␦ mRNA and protein stability. In agreement with an increase in AKR1C9 expression and conversion of tibolone to its reduced derivatives, the stimulatory effect of tibolone on ER␣-dependent transcription was also enhanced with C/EBP␦ overexpression (Fig. 6B), whereas no significant effect was evident with 17␤E, which is endogenously reduced at its 3␣ position (Fig. 6C).  1 and 1 M) and forskolin significantly enhanced AKR1C9 gene promoter activity (p Ͻ 0.05). In C, cells were treated with PGE2 for the times indicated and AKR1C9 mRNA was assessed by Northern analysis. The single panel on the right shows a DNA ladder to estimate the size of AKR1C9 mRNA, and the lower panel shows rRNA levels, visualized by staining with ethidium. 24 NOVEMBER 23, 2007 • VOLUME 282 • NUMBER 47

JOURNAL OF BIOLOGICAL CHEMISTRY 34007
C/EBP␦ Binds Consensus C/EBP Response Elements in the AKR1C9 Gene Promoter in Osteoblasts-Initial characterization studies revealed multiple C/EBP binding sites in the 2.0-kb 3Ј-region of the rat AKR1C9 gene promoter (15). Recent re-inspection of this region with MatInspector (Genomatix Software, GmbH) suggested several highly probable C/EBP response elements, as defined by significant degrees of core and matrix similarity. EMSA with oligonucleotide probes comprising three of these putative C/EBP response elements in the AKR1C9 gene promoter (Table 1) exhibited multiple nuclear factor binding complexes. A significant increase in binding by nuclear factor from PGE2-induced cells was limited to slowly migrating complexes in the upper portion of the binding profiles. The increase in nuclear factor binding seen with radiolabeled probe C1/2, which contains 2 C/EBP response elements at nucleotides Ϫ1728 to Ϫ1720 (designated as site C/EBP site C1) and Ϫ1704 to Ϫ1696 (site C2), was sensitive to competition with unlabeled oligonucleotide C1/2 or C1/2, in which only binding site C2 was mutated, but not with oligonucleotide C1/2, in which only binding site C1 was mutated (Fig. 7A, left  panel). The increase in binding seen with radiolabeled probe C3, containing a C/EBP response element at nucleotides Ϫ1380 to Ϫ1372 (site C3) was sensitive to competition with unlabeled oligonucleotide C3 (Fig. 7A, right panel). With both radiolabeled probes C1/2 and C3, the PGE2-induced nuclear protein complexes were also sensitive to competition by unlabeled probe HS3D, the C/EBP response element that occurs in exon 1 in the IGF-I gene promoter (34) (Fig. 7A) and to anti-C/EBP␦ antibody (Fig. 7B). The presence of a doublet complex is consistent with C/EBP␦ binding in other promoter DNA contexts (31,39), although the reason for more than a single band remains unclear. The lack of C/EBP␦ binding to site C2 may relate to sequence variations within the response element itself or its flanking domains, or to the possibility that site C2 may be more readily sensitive to other C/EBP isoforms not present in   Fig. 4, in combination with vector (0) or up to 150 ng/cm 2 of expression plasmids encoding C/EBP␦ (left panel), truncated dominant negative C/EBP␦ (dn-C/EBP␦) (middle panel), or antisense-oriented C/EBP␦ (␣s-C/EBP␦) as indicated. The cells were then treated for 6 h with control vehicle (con) or 1 M PGE2) as indicated. In B, cells were co-transfected for 24 h with ER␣ and reporter plasmid ERE as in Fig. 1, in combination with vector (vec) or C/EBP␦ expression plasmid at 37.5 ng/cm 2 in the right panel, or at the amounts shown in the left panel, and then treated with control vehicle (con) or tibolone as indicated. In C, cells were co-transfected ER␣ and reporter plasmid ERE in combination with vector or C/EBP␦ expression plasmid at 37.5 ng/cm 2 , and then treated with vehicle, tibolone, or 17␤E as indicated. Co-transfection with all concentrations of C/EBP␦ significantly enhanced, and co-transfection with all concentrations of dn-C/EBP␦ or ␣s-C/EBP␦ significantly suppressed the stimulatory effect of PGE2 on AKR1C9 gene promoter activity (p Ͻ 0.05). Co-transfection with C/EBP␦ significantly enhanced ER␣-dependent gene expression in response to tibolone (p Ͻ 0.05).

C3
C1/2 C3 probe C1/2 PGE2 -+ + + + + -+ + + + + + + + + . C/EBP␦ binds to select sites in the AKR1C9 gene promoter. Fetal rat osteoblasts were treated for 4 h with vehicle (0) or 1 M PGE2 as indicated, and nuclear extracts were tested by EMSA with 32 P-labeled oligonucleotide probes C1/2 and C3 from the AKR1C9 gene promoter defined by C/EBP binding domains as indicated in Table 1. In A, nuclear extracts were preincubated with no addition or 100-fold excess of unlabeled native or mutated oligonucleotides as indicated in Table 1, and then supplemented with 32 P-labeled probes. In B nuclear extracts were preincubated with non immune IgG, antibody (␣-) to C/EBP␦, or irrelevant control antibody (AB), and then supplemented with 32 P-labeled probes. Identical results were obtained with nuclear extracts from two separate studies.

TABLE 1 Oligonucleotides used in EMSA studies
Probe C1/2 comprises nucleotides Ϫ1744 to Ϫ1689, and probe C3 comprises nucleotides Ϫ1392 to Ϫ1354 from the rat AKR1C9 gene promoter (GenBank TM accession AF180326). The underlined regions indicate sequences with high homology to other published C/EBP core binding sequences ͓T-(T/G)-N-N-(T/G)-(C/T)-A-A-(T/G)͔ (where X/X indicates either of the nucleotides shown, and N equals any nucleotide), in forward or reversed orientation. In probes with the designation , nucleotide substitutions that disrupt the C/EBP binding sequence are shown in italics with no underlining. In probe HS3D, the underlined region indicates the previously identified C/EBP binding sequence found in exon I of the IGF-I gene (47).

Name Sequence
the osteoblast nuclear extracts. However, unlike C/EBP␦, overexpression of C/EBP␤ failed to increase AKR1C9 gene promoter activity (Fig. 8A), analogous to results seen with the TGF-␤ receptor III gene promoter (39). Also, anti-C/EBP␤ antibody had little or no effect on nuclear factor binding within this region (Fig. 8B). Thus, nuclear factor C/EBP␦ from PGE2 induced osteoblasts can bind specifically to at least two distinct sites in the AKR1C9 gene promoter. In addition to these sequences, at least five other possible C/EBP binding sites occur within the 2.0-kb 3Ј-region of the AKR1C9 promoter (13). This large number of C/EBP binding sites makes it especially difficult to determine the relative importance of individual, or combinations of, possible C/EBP response elements with certainty. Nonetheless, stimulation by native C/EBP␦ overexpression and inhibition by dominant negative C/EBP␦ on AKR1C9 gene promoter activity predict a clear role for C/EBP␦ in the context of select cells like osteoblasts, or to specific physiological conditions.

DISCUSSION
Sex steroids have important targets beyond reproductive tissues. For instance, bone integrity decreases when endogenous sex steroid levels fall during aging, and can be maintained in large part by sex steroid HRT. In bone forming osteoblasts, sex steroids drive gene expression directly through cis-acting DNA response elements that bind hormone activated SSRs, and indirectly through trans-acting interactions between activated SSRs and other transcriptional components or other signaling events (51,52). Sex steroids also regulate gene expression in neural, cardiovascular, skin, and other connective tissue cells, and more recent evidence reveals an increase in breast and vascular disease with long term sex HRT (51, 53-55). Therefore, there is great interest to identify SSR agonists with more function restricted or tissue selective effects (56,57).
Several compounds, while structurally similar to sex steroids, contain modifications that could enhance their stability or vary their interactions with SSRs in highly focused ways (58,59) and appear particularly sensitive to enzymatic modification. This occurs through several enzymes, first identified in liver, which can metabolize precursor compounds as well as active SSR agonists. For example, 3␣-hydroxysteroid dehydrogenases can convert the synthetic compound estren, a weak ER␣ or AR agonist, to the potent AR agonist 19-nortestosterone (12), whereas individual AKR1C family members can convert the synthetic compound tibolone, itself an AR agonist, to the potent ER␣ agonists 3␣-hydroxytibolone and/or 3␤-hydroxytibolone (17,18). Both enzyme families modify native compounds during the course of steroid synthesis and inactivation, and are also involved in bile acid, retinoid, and PG metabolism, and the carcinogenic activation of some aromatic hydrocarbons (1-3, 60, 61).
Most evidence suggests that several AKR1C gene family members act to reduce specific ketosteroids to 3␣-, 3␤-, or 20␣hydroxysteroids. In some cases, however, they can also perform the reverse, oxidative reaction. Rat AKR1C9 is one of the best studied members of this gene family. Recombinant AKR1C9 exhibits potent, bi-directional activity when supplemented with appropriate substrates and cofactors in vitro. Nonetheless, AKR1C9 is primarily reductive when overexpressed within the cellular context (1)(2)(3)(4)(5)(6), predicting that oxidation principally relies on other, perhaps short chain dehydrogenase reductase/ short chain oxidoreductase type enzymes. AKR1C9 expression has mostly been studied in liver where its levels are constitutively high and may be sensitive to control by sex steroids and glucocorticoids (62)(63)(64). Original analysis of the AKR1C9 gene promoter revealed many response elements in addition to those for steroid hormones (13)(14)(15), suggesting complex levels of regulation.
We previously reported that osteoblasts possess potent oxidative 3␣-hydroxysteroid dehydrogenase activity (12), and show here that they also exhibit significant reductive 3-ketosteroid reductase activity by which they rapidly convert the AR agonist tibolone to the ER␣ agonists 3␣-and 3␤-hydroxytibolone. 3␣-hydroxytibolone is the sole tibolone metabolite produced by recombinant rat AKR1C9 in vitro (18), unlike the appearance of both 3␣and 3␤-hydroxytibolone in intact rat osteoblasts. Thus other enzymes with 3-ketosteroid reductase activity could account for the appearance of 3␤-hydroxytibolone in rat osteoblasts, or substrate metabolism may differ in subtle but still unknown ways between in vitro assay conditions and intact cells. Even so, tibolone, by way of metabolism, can directly drive gene expression through cis-acting EREs and indirectly activate the essential osteoblast transcription factor Runx2. Little is yet known regarding native genes driven directly through ERE in osteoblasts. However, we found that the ERE-driven oxytocin gene promoter is also sensitive to agonist-dependent ER␣ activation in osteoblasts, 4   In A, fetal rat osteoblasts were transfected for 24 h with a 2.0-kb fragment of the rat AKR1C9 gene promoter cloned upstream of firefly luciferase (75 ng/cm2) in combination with vector (0) or 37.5 ng/cm2 of expression plasmids encoding C/EBP␤ or C/EBP␦. The cells were then treated for 6 h with vehicle (control) or 1 M PGE2 as indicated. Co-transfection with C/EBP␦ significantly enhanced AKR1C9 gene promoter activity and significantly increased the stimulatory effect of PGE2 (p Ͻ 0.05), whereas co-transfection with C/EBP␤ had no effect. In B, nuclear extracts from control (Ϫ) or PGE2-treated cells (ϩ) were preincubated with no addition, non-immune IgG, or antibody (␣-) to C/EBP␤ or C/EBP␦ as indicated, and then supplemented with 32 P-labeled probe C3 (Table 1). Identical results were obtained with nuclear extracts from two separate studies. NOVEMBER 23, 2007 • VOLUME 282 • NUMBER 47 effects on native estrogen-sensitive genes. Even so, indirect transcriptional effects that result from ER␣ activation may be equally if not more important physiologically. In this regard, transgenic mice generated to express a mutated ER␣ that fails to drive gene expression through ERE have multiple reproductive related tissue defects, but retain an increase in uterine cell proliferation and a decrease in luteinizing hormone secretion in response to steroid (65)(66)(67). Mice expressing a single copy of the mutated ER␣ gene also exhibit a focused deficit in cortical bone. However, they display a seemingly paradoxical increase in bone mass after ovariectomy that is suppressed by HRT (68). These and other studies clearly show that osteoblasts possess the molecular capacity to respond directly and indirectly to estrogens, and rely on a balance in both response systems.

Osteoblast 3-Ketoreductase Activity and Expression
We also found that the gene promoter for AKR1C9 is induced in osteoblasts by PKA activation and is driven by transcription factor C/EBP␦. This finding is analogous to those from analyses of insulin-like growth factor I and transforming growth factor ␤ receptor type III gene promoters, which also respond rapidly to PKA-dependent activation and translocation of pre-existing C/EBP␦ in rat and human osteoblasts (31,39,50). Original inspection of the AKR1C9 gene promoter identified multiple C/EBP response elements. However, initial analysis with a single oligonucleotide probe derived from a distal upstream region between nucleotides Ϫ4349 to Ϫ4277 of the AKR1C9 promoter with a near consensus C/EBP site at Ϫ4285 to Ϫ4277, showed no binding by recombinant C/EBP␣ or C/EBP␤, and no competition by a consensus C/EBP response element in combination with nuclear extract from human hepatoma derived HepG2 cells (15). Those results may differ from ours which focused on more downstream response elements, and suggest the possibility that not all consensus C/EBP sites are competent to bind this transcription factor, as we found here and in previous studies (39). Alternately, they may relate to preferential binding by isoform C/EBP␦, which may not have been present in the HepG2 cell nuclei, to distinct elements, or to the absence of sufficient 3Ј-terminal flanking sequences in the probe used in those initial studies (15). Curiously, glucocorticoid increases C/EBP␤ and C/EBP␣ expression in rat hepatocytes (69), where it also induces AKR1C9 gene promoter activity (14). In this instance however, the stimulatory effect of glucocorticoid relies at least in part on imperfect glucocorticoid response elements that may counteract the effect of constitutive nuclear factor occupancy at inhibitory Oct binding sites (14). In contrast, glucocorticoid increases the expression of C/EBP␤ and C/EBP␦ in osteoblasts (31), but by itself directly suppresses rather than enhances AKR1C9 gene promoter activity in these cells. However, consistent with analysis of IGF-I gene expression, transient exposure to glucocorticoid enhances C/EBP␦ expression and has a synergistic stimulatory effect on subsequent exposure to PGE2 (31) (supplemental Fig. S2). Therefore, C/EBP␦ may account for PKA-dependent expression of AKR1C9 in osteoblasts, but not in liver cells, and the direct inhibitory effect of glucocorticoid in both cell types appears unrelated to changes in C/EBP expression or activity.
Our evidence for the stimulatory effect of PGE2 on AKR1C9 expression adds the possibility for autoregulation of PG synthesis, inasmuch as members of the AKR1C gene family control the reduction of either PGH2 or PGE2 itself to PGF2␣, which is a potent inducer of cGMP and PKC in osteoblasts and regulates PGE2 synthesis (1,70,71). High levels of PGF2␣ might then limit PGE2 activity (21), perhaps accounting in part for its biphasic effects on gene expression. Furthermore, since PGE2 favors the possibility of higher local estrogen levels in the skeletal tissue environment by way of C/EBP␦ activation and AKR1C9 gene expression, the system may also be self-limiting to the extent that activated ER␣ can complex with and limit C/EBP-dependent gene expression in osteoblasts (40,72). Finally, as earlier reported, osteoblasts also express potent 3␣-hydroxysteroid dehydrogenase activity (12), which promotes substrate oxidation. Therefore, an eventual restriction in AKR1C9 expression could enhance the androgenic potential of certain oxidoreductase-sensitive substrates. In this regard we found that antisense suppression of AKR1C9 expression by osteoblasts significantly enhanced AR-dependent gene expression by tibolone, DHT, and estren (supplemental Fig. S3), consistent with possible counteracting effects by these enzyme families.
In summary, our studies show that osteoblasts possess endogenous 3-ketosteroid reductase activity that allows them to reduce facile substrates like tibolone to potent ER␣ agonists,  . Control of steroid precursor metabolism and AKR1C9 expression in osteoblasts. In rat osteoblasts, the 3-ketosteroid reductase AKR1C9 can reduce facile 3-ketosteroids like tibolone to 3-hydroxysteroids with strong ER␣ activating potential. In these cells, AKR1C9 gene expression can be induced by hormones like PGE2 that increase cAMP, release the active PKA catalytic subunit, and enhance C/EBP␦ activation and nuclear transport. Ligand-activated ER␣ may then induce direct transcriptional effects through ERE, and indirect effects through the osteoblast-enriched transcription factor Runx2 and through Smads in response to TGF-␤ stimulation. and perhaps to limit endogenous local androgen levels in the skeleton. We found that the 3-ketosteroid reductase activity can be accounted for in large part by the AKR gene family member AKR1C9 in rat osteoblasts, where its expression is induced by PGE2 in a PKA-dependent way through activation of C/EBP␦. Ligand-activated ER␣ may then induce direct transcriptional effects through ERE, and indirect effects through Runx2 and Smads in response to TGF-␤, as modeled in Fig. 9. Importantly, the balance between steroid and steroid precursor reduction and oxidation, through differences in the local redox state or changes in the relative expression of endogenous AKR and short chain reductase dehydrogenase/short chain oxidoreductase enzyme family members, could significantly affect specific SSR agonists and impact bone integrity, and may be more readily apparent with selective HRT. Further studies will help to define the importance of C/EBP␦ on AKR expression and activity during bone remodeling, mechanical load, trauma, and inflammatory disease, where activation of PKA drives changes in C/EBP␦ expression and activity in osteoblasts and in this way favors ketosteroid substrate reduction and an increase in the level of ER agonists.