Novel Aptamer-Based Small-Molecule Drug Screening Assay to Identify Potential Sclerostin Inhibitors against Osteoporosis

A novel aptamer-based competitive drug screening platform for osteoporosis was devised in which fluorescence-labeled, sclerostin-specific aptamers compete with compounds from selected chemical libraries for the binding of immobilized recombinant human sclerostin to achieve high-throughput screening for potential small-molecule sclerostin inhibitors and to facilitate drug repurposing and drug discovery. Of the 96 selected inhibitors and FDA-approved drugs, six were shown to result in a significant decrease in the fluorescence intensity of the aptamer, suggesting a higher affinity toward sclerostin compared with that of the aptamer. The targets of these potential sclerostin inhibitors were correlated to lipid or bone metabolism, and several of the compounds have already been shown to be potential osteogenic activators, indicating that the aptamer-based competitive drug screening assay offered a potentially reliable strategy for the discovery of target-specific new drugs. The six potential sclerostin inhibitors suppressed the level of both intracellular and/or extracellular sclerostin in mouse osteocyte IDG-SW3 and increased alkaline phosphatase activity in IDG-SW3 cells, human bone marrow-derived mesenchymal stem cells and human fetal osteoblasts hFOB1.19. Potential small-molecule drug candidates obtained in this study are expected to provide new therapeutics for osteoporosis as well as insights into the structure–activity relationship of sclerostin inhibitors for rational drug design.


Introduction
Osteoporosis is considered a major and global public health concern. Because osteoporosis occurs primarily as a result of normal ageing and menopausal estrogen deficiency, of the hammerhead ribozyme and an anti-Rev aptamer [24]. A structurally diverse library of antibiotics was screened with this platform for novel inhibitors of HIV replication. The binding affinity of aptamers is usually unaffected by conjugation with detection molecules such as fluorescent dyes, and labeling is not necessary for the target protein or the smallmolecule drug candidates. These aptamer-based assay systems therefore offer new and powerful molecular biology tools for target validation and lead identification in drug discovery [22,25].
In this study, an aptamer-based competitive assay was established to screen for potential small-molecule sclerostin inhibitors and to facilitate drug repurposing by producing reliable hits from currently available chemical libraries. Drug repurposing, the process of discovering new therapeutic uses for existing drug candidates, is considered to shorten the time required to identify new drug candidates and increase the productivity of the pharmaceutical industry [26]. As low-molecular-weight, synthetic chemicals dominate the pharmaceutical market, the vast amount of chemicals synthesized and studied has culminated to various chemical libraries and databases. The development of efficient HTS assays has therefore become a crucial step to accelerate drug repurposing and drug discovery. Because aptamers are characterized by specific binding and antagonistic effects on protein function [25], sclerostin is a suitable target for competitive drug screening using sclerostin-specific aptamers. Chemical compounds with affinity comparable to that of an aptamer are expected to be potent inhibitors of sclerostin. The idea of a competitive assay is not new, but has not been fully explored in aptamer-based drug screening. We propose to establish an aptamer-based competitive assay in which the fluorescence-labeled, sclerostin-specific aptamers compete with compounds from selected chemical libraries for binding to immobilized recombinant human sclerostin to achieve high-throughput screening for potential small-molecule sclerostin inhibitors.

Construction of the Chemical Library for Aptamer-Based Drug Screening
Studies on biomaterials designed to induce osteogenic differentiation found that the functional groups exposed to human mesenchymal stem cells may control their differentiation fate [6,[27][28][29]. Although an absolute correlation has not been concluded, it was shown that the amino group is one of the functional groups to promote osteogenic differentiation [6,29]. Compounds with tertiary amino groups may be protonated at neutral pH of the cell culture medium to provide a biochemical and electrical environment favorable to osteogenesis. As our first attempt in drug screening for sclerostin inhibitors, 96 compounds containing primary, secondary or tertiary amino groups were selected from the commercially available inhibitor library and FDA-approved drug library. Of the 96 selected chemicals, 40 are FDA-approved drugs, 42 contain at least one tertiary amino group, 35 have primary and/or secondary amino groups but not tertiary ones, and 19 exhibit no amino group (Supplementary Table S1).

Aptamer-Based Competitive Drug Screening Assay
Because the amino acid sequence of mouse sclerostin and human sclerostin share 86.9% identity (Figure 1), a DNA aptamer against mouse sclerostin identified by Shum et al. was applied in the design of the aptamer-based competitive drug screening assay [30]. As shown in Figure 2a, immobilization of sclerostin in the 96-well black plate relied on the binding affinity between the nickel (Ni 2+ ) coating on the plate surface and the His-tagged recombinant human sclerostin protein. The Ni 2+ -coated plate was pre-blocked with bovine serum albumin (BSA), allowing for the elimination of BSA blocking after the binding reaction. Ni 2+ -based affinity binding is sensitive to metal chelating agents such as EDTA and reducing agents such as mercaptoethanol, but is compatible with nonionic detergents and dimethyl sulfoxide (DMSO), in which stock solutions of the chemical compounds were prepared. The density of immobilized sclerostin protein on the Ni 2+ -coated surface was determined according to the manufacturer's instruction, which indicates that the binding capacity of the nickel-coated 96-well plate is~9 pmol/well, corresponding tõ 200 ng/well His-tagged sclerostin protein. The immobilized human sclerostin was then reacted with 500 nM FAM-labeled sclerostin-specific aptamer in the presence of potential drug candidates.  Aptamer-based and antibody-based competitive assays for small-molecule drug screening. The assay platforms were constructed by immobilizing His-tagged recombinant human sclerostin on a nickel-coated surface, followed by reacting with small-molecule compounds as potential drug candidates in the presence of (a) FAM-labeled sclerostin-specific aptamer or (b) anti-sclerostin antibody followed by detection with FITC-labeled secondary antibodies.
It is assumed that compounds with binding affinity close to or higher than that of the aptamer may associate with the target protein, thus displacing the fluorescence-labeled aptamer. Potential inhibitors of sclerostin can then be identified by a significant decrease in fluorescence compared to that of the aptamer-only control. In the presence of 20 µM small-molecule inhibitors, 6 of the 96 selected inhibitors resulted in a significant decrease in fluorescence to more than 50% of the aptamer-only control ( Figure 3a and Table 1). Results from the aptamer-based drug screening were compared to those of an antibody-based competitive immunoassay, in which the six selected small-molecule inhibitors competed with anti-sclerostin antibodies for binding to the recombinant sclerostin protein (Figure 2b). A higher concentration (200 µM) of the chemical library was used in the antibody-based competitive drug screening because at 20 µM, the reduction in fluorescence caused by the six potential sclerostin inhibitors was insignificant. As shown in Figure 3b, two of the six chemical compounds listed in Table 1, baicalein and alizarin, were found to significantly reduce the fluorescence intensity of the sclerostin antigen-antibody immunocomplex, although the decrease in fluorescence (40 and 55% relative to the control for baicalein and alizarin, respectively) was less significant than that observed in the aptamer-based competitive drug screening assay (88 and 86% relative to the control for baicalein and alizarin, respectively). The less significant reduction in fluorescence and the higher concentration of the chemical library required in the antibody-based competitive assay may be explained at least in part by the higher affinity of the antibody (K D usually in the nanomolar range) toward the sclerostin protein, compared to that of the aptamer (K D = 0.67 ± 0.08 µM according to [30]). The lower molecular weight of the 30-mer DNA aptamer (9.5 kDa) compared to the 150-kDa anti-sclerostin antibody may also lead to easier replacement by small-molecule compounds, making aptamers a more suitable "competitor" in low-molecular-weight drug screening. Representative results of six other chemicals which did not cause over 50% reduction in fluorescence in the aptamer-based competitive drug screening assay are shown in Figure 3c for comparison, with the corresponding results from the antibody-based competitive assay in Figure 3d. Fluorescence intensity of the sclerostin-specific aptamer or the sclerostin antigen-antibody immunocomplex in the presence of selected small-molecule inhibitors. His-tagged sclerostin protein was immobilized at 200 ng/well in a nickel-coated 96-well black plate and reacted with (a,c) 500 nM FAM-labeled sclerostin-specific aptamer in the presence or absence of 20 µM of one of the inhibitors, or (b,d) 0.52 µg/mL anti-sclerostin antibody (dilution ratio 1:500) in the presence or absence of 200 µM of one of the inhibitors, followed by detection with FITC-labeled secondary antibodies. Fluorescence intensity was determined by a TECAN Infinite 200 PRO F Nano + microplate reader at Ex485 nm/Em520 nm. Error bars represent standard deviations (n ≥ 3). * p < 0.05 and ** p < 0.01, compared to the control. Table 1. A list of potential drug candidates screened from 96 small-molecule inhibitors that resulted in more than 50% reduction in fluorescence intensity compared to the control in the aptamer-based competitive drug screening assay. We then investigated whether fluorescence quenching was responsible for the reduction of fluorescence observed in Figure 3a,b.The aptamer-based competitive drug screening assay was performed as described previously but without the immobilized sclerostin. When 20 or 200 µM of the chemical compound was incubated with 500 nM FAM-labeled sclerostin aptamer, a slight decrease in fluorescence intensity was observed for baicalein ( Figure 4). Although alizarin absorbs at the emission wavelength of FAM (520 nm), the quenching effect of alizarin was not significant at both 20 and 200 µM. On the other hand, avasimibe enhanced the florescence of FAM at both 20 and 200 µM. Nevertheless, these conditions are not comparable to the actual experimental setting during fluorescence detection. When competitive drug screening was performed in the presence of immobilized sclerostin, the unbound aptamer and chemical compound were washed away prior to fluorescence detection. Unless the chemical compound was tightly associated with or in close proximity to the FAM-labeled aptamer, the amount of residual baicalein or alizarin in the final solution should be much lower than 20 µM for fluorescence quenching to occur. Furthermore, in the antibody-based competitive drug screening assay, unbound anti-sclerostin antibodies and the small-molecule inhibitors were washed away before the FITC-labeled secondary antibody was added, at which stage the residual chemical compound should be bound to sclerostin ( Figure 2b). As the possibility of interaction between unbound baicalein or alizarin and the FITC-labeled secondary antibody is relatively unlikely, the reduction in fluorescence can be attributed primarily to the displacement of anti-sclerostin antibodies by baicalein or alizarin.

Intracellular and Extracellular Levels of Sclerostin in Mouse IDG-SW3 Cells
Physiologically, sclerostin is secreted by mature, embedded osteocytes, but according to the literature, existing osteocyte cell lines lack the expression of sclerostin [31]. In order to study the effect of the drug candidates on sclerostin expression, the mouse osteoblastto-late-osteocyte cell line IDG-SW3 was used [31]. As shown in Figure 5, when cultured in osteogenic induction medium (OIM) for 21 days, the level of intracellular sclerostin was increased to approximately three-fold of the control, while that of extracellular sclerostin remained similar to that of the control. In the presence of 0.001-1 µM of the six potential sclerostin inhibitors, intracellular sclerostin was almost depleted, while extracellular sclerostin decreased significantly at several or all concentrations tested compared to the untreated and OIM controls, especially for avasimibe, alizarin, baicalein and D4476. Although sclerostin was proven to be a crucial therapeutic target for skeletal diseases, its molecular mechanism of regulation and degradation remains unclear. In addition to transcriptional regulation through the sclerostin gene, a recent report has shown that rapid, minute-scale lysosomal degradation may also be responsible for its down-regulation [32]. Nevertheless, these results provide strong evidence that the six potential sclerostin-binding inhibitors effectively suppressed the level of both intracellular and extracellular sclerostin. Results for two other chemical compounds, GSK256066 and voriconazole, which did not cause a significant reduction in fluorescence in the aptamer-based and antibody-based competitive drug screening assays, as shown in Figure 3c,d, are also presented in Figure 5 for comparison. Voriconazole, an inhibitor of the P450 isoenzyme CYP2B6 and an antifungal medication, did not affect the protein expression of sclerostin intracellularly or extracellularly. In contrast, intracellular and extracellular levels of sclerostin were both significantly reduced in the presence of GSK256066, a phosphodiesterase 4 inhibitor. IDG-SW3 cells were treated with 0.001-1 µM of one of the 6 potential sclerostin inhibitors, as well as the negative controls GSK256066 and voriconazole, in OIM for 21 days, followed by collection of the cell lysate and conditioned medium to determine the (a) intracellular and (b) extracellular sclerostin concentration, respectively, with a mouse sclerostin ELISA kit. The concentration of sclerostin was expressed as the amount of sclerostin (pg) normalized to total protein (mg) in the cell lysate or conditioned medium. Error bars represent standard deviations (n ≥ 3). * p < 0.05 and ** p < 0.01 compared to the OIM controls.

Cellular Rescue Assay and Osteogenic Activity of Potential Sclerostin-Binding Inhibitors
To examine whether the sclerostin-binding effect of potential sclerostin inhibitors observed in the drug screening assay was valid at the cellular level, a rescue assay was designed by pre-incubating recombinant human sclerostin (rhSOST) with anti-sclerostin rabbit antibody or selected small-molecule inhibitors under osteogenic induction conditions at room temperature for 30 min prior to culturing with human bone marrow-derived mesenchymal stem cells (hBMSCs) for three days. After culturing in OIM for another four days, the result of the rescue assay was assessed by measuring alkaline phosphatase (ALP) activity, an early marker for osteoblast maturation. As shown in Figure 6, ALP activity was completely abrogated in the presence of 10 µg/mL rhSOST (rhSOST vs. control and OIM), but remained unaffected at 1 µg/mL rhSOST (data not shown). Treatment with 10 µg/mL anti-sclerostin antibody did not further enhance ALP activity because secreted sclerostin is undetectable in vitro in hBMSC culture in the early stage of osteogenic differentiation for sclerostin neutralization to take place (anti-SOST vs. OIM in Figure 6). However, in the presence of exogenous rhSOST, the suppressed ALP activity was partially alleviated by co-incubating with the anti-sclerostin antibody (rhSOST + anti-SOST vs. control and OIM in Figure 6). The partial recovery in ALP activity was also observed when small-molecule inhibitors such as avasimibe, baicalein, alizarin, evacetrapib or GSK256066 were present ( Figure 6). Figure 6. Effect of the potential sclerostin inhibitors on rescuing suppressed ALP activity caused by exogenous sclerostin. Recombinant human sclerostin (rhSOST) was incubated with 1 µM of the chemical compound for 30 min prior to incubation with hBMSCs for 3 days. After refreshing the medium with OIM, the cells were cultured for another 4 days before ALP activity was determined. Error bars represent standard deviations (n ≥ 3). ## p < 0.01, compared to the control; ** p < 0.01, compared to the OIM control.
To further demonstrate whether the ALP activity was enhanced by potential sclerostinbinding inhibitors in the rescue assay, hBMSCs were cultured for 21 days in OIM containing the inhibitors to analyze their osteogenic activity. Enhanced ALP activity was found at 0.1 µM atorvastatin, baicalein or evacetrapib and at 0.01 µM alizarin or D4476 in hBM-SCs (Figure 7). In comparison, no significant change in ALP activity was observed in the presence of voriconazole, while GSK256066 down-regulated ALP activity in a dosedependent manner. When IDG-SW3 was cultured in OIM with baicalein and alizarin, the two chemical compounds that resulted in the most significant decrease in fluorescence in both aptamer-based and antibody-based drug screening, the ALP activity of the mouse osteocyte was significantly higher than that of the control and OIM-treated cells at 10 nM baicalein, but decreased with increasing concentrations of baicalein, while that in the presence of alizarin increased dose-dependently (Figure 8a,b). On the other hand, higher ALP activity was detected in hFOB1.19 cells compared to the OIM controls for both baicalein and alizarin (Figure 8c,d). As opposed to sclerostin-specific aptamer and antibody, whose effect was mediated extracellularly through neutralization of secreted sclerostin, the small-molecule inhibitors may enter the cells and exert their effect both intracellularly and extracellularly. The accelerated osteogenic differentiation observed in IDG-SW3 cells in the presence of baicalein or alizarin may be associated with the suppression of intracellular and extracellular sclerostin ( Figure 5). However, as the protein expression of sclerostin was barely detectable in hBMSCs and hFOB1.19 cells (data not shown), due possibly to the late expression of sclerostin in the osteogenic differentiation process of osteoblast to osteocyte, the direct association of sclerostin suppression to osteogenic differentiation in hBMSCs and hFOB1.19 cells remains to be investigated. The enhanced ALP activity observed in hBMSCs and hFOB1.19 therefore suggests that in addition to interacting directly with intracellular and extracellular sclerostin, the potential sclerostin inhibitors may act through other pathways as well to stimulate bone formation. The cellular assays, including the analysis of sclerostin protein level ( Figure 5) and ALP activity assay (Figures 6-8), may serve as preliminary in vitro validations to assess the osteogenic potential of the drug candidates in relation to sclerostin. The level of alkaline phosphatase activity was quantitated by solubilizing the cells in DMSO and determining the optical density at 570 nm, which was then normalized to total protein (mg). Error bars represent standard deviations (n ≥ 3). # p < 0.05 and ## p < 0.01, compared to the control; * p < 0.05 and ** p < 0.01, compared to the OIM controls.  The level of alkaline phosphatase activity was quantitated by solubilizing the cells in DMSO and determining the optical density at 570 nm, which was then normalized to total protein (mg). Error bars represent standard deviations (n ≥ 3). # p < 0.05 and ## p < 0.01, compared to the control; * p < 0.05 and ** p < 0.01, compared to the OIM controls.

Discussion
The small-molecule inhibitors were initially selected from the chemical libraries and categorized based on the amino group, because studies have shown that it is one of the functional groups to promote osteogenic differentiation. The functional group served only as an initial criterion for selecting compounds from the vast database of the chemical libraries, and a correlation between the amino group and the affinity to sclerostin was not assumed or observed in our drug screening results. Among the six potential sclerostin inhibitors, atorvastatin contains a secondary and a tertiary amino group, evacetrapib has two tertiary amino groups and D4476 has a primary and a secondary amino group, while no amino group is present in avasimibe, baicalein and alizarin. Although the binding site of the sclerostin-specific aptamer or antibody remains unclear, the NMR structure of both human and mouse sclerostin revealed surface areas populated with positively charged amino acids to interact with the negatively charged heparin or the negatively charged region of human low-density lipoprotein receptor-related protein 5 (LRP5), respectively [33,34]. The highly positively charged surface of sclerostin may not favor the interaction of compounds with protonated tertiary amino groups, but could provide binding sites for those with carboxyl groups such as atorvastatin and evacetrapib. In addition, the hydrophobic patch found on human sclerostin may be considered another site for the binding of relatively nonpolar or uncharged ligands [33].
Because the conditions for competitive drug screening do not correspond to the actual physiological condition, we analyzed the effect of the potential sclerostin inhibitors on cellular sclerostin level and ALP activity as further validation. Although it was as-  . (b,d) The level of alkaline phosphatase activity was quantitated by solubilizing the cells in DMSO and determining the optical density at 570 nm, which was then normalized to total protein (mg). Error bars represent standard deviations (n ≥ 3). # p < 0.05 and ## p < 0.01, compared to the control; * p < 0.05 and ** p < 0.01, compared to the OIM controls.

Discussion
The small-molecule inhibitors were initially selected from the chemical libraries and categorized based on the amino group, because studies have shown that it is one of the functional groups to promote osteogenic differentiation. The functional group served only as an initial criterion for selecting compounds from the vast database of the chemical libraries, and a correlation between the amino group and the affinity to sclerostin was not assumed or observed in our drug screening results. Among the six potential sclerostin inhibitors, atorvastatin contains a secondary and a tertiary amino group, evacetrapib has two tertiary amino groups and D4476 has a primary and a secondary amino group, while no amino group is present in avasimibe, baicalein and alizarin. Although the binding site of the sclerostin-specific aptamer or antibody remains unclear, the NMR structure of both human and mouse sclerostin revealed surface areas populated with positively charged amino acids to interact with the negatively charged heparin or the negatively charged region of human low-density lipoprotein receptor-related protein 5 (LRP5), respectively [33,34]. The highly positively charged surface of sclerostin may not favor the interaction of compounds with protonated tertiary amino groups, but could provide binding sites for those with carboxyl groups such as atorvastatin and evacetrapib. In addition, the hydrophobic patch found on human sclerostin may be considered another site for the binding of relatively nonpolar or uncharged ligands [33].
Because the conditions for competitive drug screening do not correspond to the actual physiological condition, we analyzed the effect of the potential sclerostin inhibitors on cellular sclerostin level and ALP activity as further validation. Although it was assumed that chemical compounds which compete with the aptamer or antibody for binding to sclerostin may be considered potential sclerostin inhibitors, this does not necessarily imply that those that do not interact directly with sclerostin have no effect on its expression or exhibit no osteogenic activity, as compounds selected for drug screening are known protein or enzyme inhibitors and may act on various pathways to regulate sclerostin and the Wnt signaling pathways. For the two selected negative controls, voriconazole did not affect the protein expression of sclerostin intracellularly or extracellularly, and no significant change in the ALP activity of hBMSCs was observed in the presence of voriconazole. In contrast, intracellular and extracellular levels of sclerostin were both significantly reduced, but ALP activity was also down-regulated in a dose-dependent manner in the presence of GSK256066. Furthermore, the results from this study indicate that the fluorescence of FAMlabeled sclerostin-specific aptamer and FITC-conjugated anti-sclerostin antibody was not affected in the presence of icariin, but the small-molecule inhibitor was reported to promote the osteogenic differentiation of rat BMSCs through the regulation of sclerostin [35].
In the cellular assays, the sclerostin-specific DNA aptamer is susceptible to nuclease degradation in cell culture medium and has to be modified in order to enhance its stability [30], and was therefore not considered as the positive control. Despite the general understanding that the anti-sclerostin antibody promotes bone formation by neutralizing secreted sclerostin, it is not a suitable positive control either because the sclerostin protein bound by the antibody remained in the conditioned medium and can still be detected in ELISA by the immobilized capture antibody that recognizes a different epitope. In addition, the osteogenic effect of the anti-sclerostin antibody, although demonstrated in animal and clinical studies, cannot be easily observed in vitro in cells induced to differentiate osteogenically. In IDG-SW3 cells, which exhibit osteoblast to late osteocyte characteristics during osteogenic differentiation, both sclerostin expression and ALP activity were enhanced, according to the literature [31] and results from this study (Figures 5a and 8a,b). It is therefore assumed that the osteogenic inducers, i.e., ascorbic acid and β-glycerophosphate supplemented in the osteogenic induction medium, overcome the antagonistic effect of secreted sclerostin, resulting in the up-regulation of ALP activity. The anti-sclerostin antibody added to IDG-SW3 cells may be continuously counteracted by secreted sclerostin, rendering its bone-stimulating effect insignificant. Moreover, the practice of refreshing the medium every three or four days may also eliminate in part the effect of accumulated sclerostin and anti-sclerostin antibodies. On the other hand, the amount of secreted sclerostin from hFOB1.19 and hBMSCs cultured under osteogenic conditions was too low for the anti-sclerostin antibody to exert its effect. It was reported that in hFOB1.19 cells, the extracellular sclerostin level was approximately 0.4 pg/mL [36], far lower than that detected in IDG-SW3 cells (~22 pg/mL in this study). For human mesenchymal stem cells, SOST gene expression was increased when cultured under osteogenesis-inducing conditions [37]. However, the extracellular level of the sclerostin protein was barely detectable in our study (data not shown).
In addition to its antagonistic effect on osteogenic differentiation, sclerostin is involved in the crosstalk between adipose and bone tissues [38]. Sclerostin promotes adipogenesis in mouse adipocyte 3T3-L1 and mouse BMSCs [39,40]. White adipose tissue was reduced in Sost −/− mice, accompanied by an increase in bone mass, which can be explained by enhanced Wnt/β-catenin signaling in the absence of sclerostin, leading to the up-regulation of fatty acid catabolism and down-regulation of fatty acid synthesis [41]. It is therefore reasonable that several of the potential sclerostin inhibitors obtained from the drug screening assay are themselves inhibitors of enzymes involved in lipid metabolism. Of the six potential sclerostin inhibitors, atorvastatin calcium, an FDA-approved drug to lower lowdensity lipoprotein cholesterol in blood, is the inhibitor of HMG-CoA reductase, avasimibe, baicalein and alizarin are cytochrome P450 inhibitors, evacetrapib (LY2484595) inhibits cholesteryl ester transfer protein (CETP), while D4476 inhibits both casein kinase 1 (CK1) and TGFβ type I receptor kinase (ALK5). Both in vitro and in vivo studies support that statins, or HMG-CoA reductase inhibitors, promote osteogenesis and bone formation [42]. Atorvaststin was proven to enhance bone formation in apolipoprotein E-deficient (apoE -/-) mice by increasing the gene and protein expression of Sirt1 and Runx2 [43]. Although its association with bone formation was unclear, treatment of adipocytes with avasimibe resulted in significant reduction of lipid accumulation, which is the reciprocal effect of osteogenic differentiation in mesenchymal stem cells [44,45]. The osteogenic potential of baicalein has been reported in human periodontal ligament cells [46], but correlation to sclerostin has not been reported. Baicalein was also shown to increase callus density and cortical width, and may be beneficial to a later stage of the bone healing process [47]. Alizarin was generally used as the anthraquinone dye in matrix mineralization staining, but its osteogenic effect has not been discussed. Nevertheless, other anthraquinone compounds such as aloin and emodin, both of which are natural anthraquinones derived from plants, were shown to induce osteogenic differentiation [48,49]. Finally, targets of evacetrapib and D4476, CETP and ALK5 are related to lipid metabolism and osteogenesis, and inhibition of which led to an increased level of high-density lipoprotein cholesterol in clinical trials [50] or enhanced alkaline phosphatase activity in osteoblastic mesenchymal cells [51], respectively. As the targets of these potential sclerostin inhibitors all play important roles in lipid or bone metabolism, and several of the compounds have already been shown to be potential osteogenic activators, it is suggested that the aptamer-based competitive drug screening assay designed in this study offered a potentially reliable strategy for the discovery of target-specific new drugs.

Materials
Sclerostin-specific aptamer, 5 -TTGCGCGTTAATTGGGGGGGTGGGTGGGTT-3 [30], labeled with 6-carboxyfluorescein (FAM) at the 3 end, was synthesized by Integrated DNA Technologies, Inc, Coralville, IA, USA. N-terminally histidine (His)-tagged recombinant human sclerostin protein and rabbit anti-sclerostin antibody was provided by Abcam, Cambridge, MA, USA. Pierce nickel-coated 96-well black plate pre-blocked with BSA was purchased from Thermo Fisher Scientific, Waltham, MA, USA. FITC-labeled rabbit secondary antibody was manufactured by Santa Cruz Biotechnology, Dallas, TX, USA. A chemical library of selected small-molecule inhibitors was provided in 10 mM stock solutions in DMSO by Selleck Chemicals, Houston, TX, USA.

Aptamer-Based Competitive Drug Screening Assay
A working solution of His-tagged recombinant human sclerostin protein was freshly prepared at a concentration of 2 ng/µL in tris-buffered saline (TBS, 50 mM Tris-Cl, pH 7.5, and 150 mM NaCl) and was added at 100 µL/well to a nickel-coated 96-well black plate. After reaction at room temperature for 1 h, each well was washed 3 times with 200 µL/well TBS buffer. To perform the competitive assay, 500 nM FAM-labeled sclerostin-specific aptamer was prepared in TBS buffer and was heated at 95 • C for 5 min, followed by cooling to room temperature for 15 min to promote the formation of 3D structures. The FAM-labeled aptamer was added to the 96-well plate at 100 µL/well, followed by mixing with one selected compound per well from the chemical library to a final concentration of 20 µM. The mixture of FAM-labeled aptamer and selected chemical compound was then incubated with the immobilized sclerostin protein at room temperature for another hour. After washing 3 times with 200 µL/well TBS buffer, the fluorescence intensity of each well was determined by a TECAN Infinite 200 PRO F Nano + microplate reader (Tecan Trading AG, Switzerland) at Ex485 nm/Em520 nm.

Antibody-Based Competitive Drug Screening Assay
As a comparative study, an anti-sclerostin rabbit antibody was substituted for the sclerostin-specific aptamer in the competitive assay. After immobilization of the Histagged recombinant human sclerostin protein at 200 ng/well on the nickel-coated surface, 0.52 µg/mL anti-sclerostin rabbit antibody (dilution ratio 1:500) prepared in TBS buffer along with 200 µM of the selected chemical compound were added in a total volume of 100 µL/well. After reaction for 1 h at room temperature followed by washing 3 times with 200 µL/well TBS buffer, each well was incubated with 100 µL FITC-labeled rabbit secondary antibody (dilution ratio 1:100) for another 1 h. After washing 3 times with 200 µL/well TBS buffer, the fluorescence intensity was determined by a TECAN Infinite 200 PRO F Nano + microplate reader at Ex485 nm/Em520 nm.

hFOB1.19 Cell Culture
The human fetal osteoblast cell line hFOB1.19 was provided by the Bioresource Collection and Research Center of the Food Industry Research and Development Institute, Hsinchu, Taiwan. Cells were maintained at 33.5 • C in the presence of 5% CO 2 in a growth medium consisting of a 1:1 mixture of DMEM and Ham's F-12 medium in the absence of phenol red (Thermo Fisher Scientific, Waltham, MA, USA), with additional 2.5 mM L-glutamine, 0.3 mg/mL G418 and 10% FBS. Osteogenic induction was performed at 37 • C in the growth medium supplemented with 5 × 10 -3 mol/L β-glycerophosphate, 0.1 g/L ascorbic acid, 10 -8 mol/L menadione and 10 -7 mol/L 1,25-dihydroxyvitamin D 3 .

SOST ELISA
IDG-SW3 cells were treated with 0.001, 0.01, 0.1 or 1 µM of a potential sclerostin inhibitor under osteogenesis-inducing conditions for 21 days. The conditioned medium and cell lysate at day 21 were collected, and the concentration of intracellular and extracellular sclerostin was determined by a Mouse/Rat SOST/Sclerostin Quantikine ELISA Kit produced by R&D Systems, Minneapolis, MN, USA. The concentration of sclerostin was then normalized to the amount of total protein, which was quantitated by a Bio-Rad DC Protein Assay kit (Bio-Rad Laboratories, Inc., Hercules, CA, USA).

Cellular Rescue Assay
Human BMSCs were seeded at 3000 cells/well in a 96-well black plate with a flat clear bottom and cultured in growth medium for 3 days. To perform the cellular rescue assay, 10 µg/mL recombinant human sclerostin protein was pre-incubated with 10 µg/mL antisclerostin rabbit antibody or 1 µM of the chemical compound for 30 min prior to culturing with hBMSCs for 3 days. After refreshing the medium with OIM, the cells were cultured for another 4 days before ALP activity was determined by an Alkaline Phosphatase Activity Fluorometric Assay Kit manufactured by BioVision Inc., Milpitas, CA, USA.

Alkaline Phosphatase Activity Assay
Human BMSCs, IDG-SW3 and hFOB1.19 were seeded at 10 4 , 8 × 10 4 and 10 4 cells/well, respectively, in a 24-well plate and cultured in the presence of various concentrations of the potential sclerostin inhibitor prepared in OIM for 14 or 21 days. ALP activity was determined by fixing the cells with 4% paraformaldehyde in PBS for 15 min, followed by washing successively with PBS and ddH 2 O. The cells were then reacted with 300 µL/well Pierce 1-Step NBT/BCIP Solution (Thermo Fisher Scientific, Waltham, MA, USA) for approximately 30 min at room temperature so that the blue stain of the ALP reaction product can be seen with the naked eye. After washing again with ddH 2 O, the stained cells were air-dried before microscopic images were captured.

Statistical Analysis
Statistical significance was evaluated by Student's t test. Levels of significance were expressed as significant, p < 0.05 and highly significant, p < 0.01, respectively.

Conclusions
Although romosozumab was recently approved as an effective therapeutic antibody drug for osteoporosis, the long-term health risk of antibody treatment remains to be evaluated. For a promising disease-related target such as sclerostin, alternative interventions such as small-molecule drugs are worth investigating, which may provide a broader selection of medication with various routes of administration to meet the diverse needs of patients with different degrees of bone loss. The common features in the chemical structure of baicalein and alizarin and their osteogenic potential demonstrated in this study suggest that flavonoids and anthraquinones such as baicalein and alizarin may serve as lead compounds for the development of bone-stimulating new drugs. With novel molecular biology tools such as aptamers, the process of lead identification is expected to be simplified and accelerated to find new cures for devastating diseases with limited therapeutic options, such as osteoporosis. Potential small-molecule drug candidates obtained in this study are expected to accelerate the development of new therapeutics as well as offer more insights into the structure-activity relationship of sclerostin inhibitors for rational drug design.