Defining Key Structural Determinants for the Pro ... - ACS Publications

Article pubs.acs.org/jnp

Defining Key Structural Determinants for the Pro-osteogenic Activity of Flavonoids Stephen Swioklo,† Kimberly A. Watson,‡ Elizabeth M. Williamson,† Jonathan A. Farrimond,§ Sophie E. Putnam,§ and Katrina A. Bicknell*,† †

Reading School of Pharmacy, University of Reading, Whiteknights, Reading RG6 6UB, U.K. School of Biological Sciences, University of Reading, Whiteknights, Reading RG6 6AJ, U.K. § GlaxoSmithKline, GSK House, 980 Great West Road, Brentford, Middlesex TW8 9GS, U.K. ‡

S Supporting Information *

ABSTRACT: Epidemiological studies suggest that fruits and vegetables may play a role in promoting bone growth and preventing age-related bone loss, attributable, at least in part, to phytochemicals such as flavonoids stimulating osteoblastogenesis. Through systematically screening the effect of flavonoids on the osteogenic differentiation of human mesenchymal stem cells in vitro and correlating activity with chemical structure using comparative molecular field analysis, we have successfully identified important structural features that relate to their activity, as well as reliably predicted the activity of compounds with unknown activity. Contour maps emphasized the importance of electronegativity, steric bulk, and a 2-C−3-C double bond at the flavonoid C-ring, as well as overall electropositivity and reduced steric bulk at the flavonoid Bring. These results support a role for certain flavonoids in promoting osteogenic differentiation, thus their potential for preventing skeletal deterioration, as well as providing a foundation for the lead optimization of novel bone anabolics.

B

complete recovery of ovariectomy-induced bone loss was observed following rutin (quercetin-3-rutinoside) supplementation in ovariectomized (OVX)-rats15 and quercetin in OVXmice,16 the latter of these studies associating the outcome predominantly with quercetin promoting bone formation. Osteogenic transplant studies also highlight the bone-anabolic properties of quercetin; promoting engraftment and bone formation in calvarial defect models.10,17 Similarly, kaempferol has been demonstrated to protect against OVX-induced bone loss in rats, both by reducing the rate of skeletal turnover and by increasing bone-forming capacity, with a greater number of osteoprogenitor cells in bone marrow examined ex vivo.13 Further to this, kaempferol injection into the calvarial periosteum of newborn rats has been shown to promote osteogenesis.18 These studies are further supported by the bone anabolic effects of kaempferol- and quercetin-rich Ginkgo biloba in rat models of OVX- and glucocorticoid-induced osteoporosis.19−21 Collectively, a convincing argument exists for kaempferol and quercetin’s potential to support skeletal homeostasis via antiosteoclastogenic and pro-osteogenic effects. Less attention however has focused on other flavonols, of which there are over 600.22

one remodeling is a highly dynamic process with approximately 5% of cortical and 20% of trabecular bone turned over annually.1 Skeletal integrity is dependent on the balance between osteoclastic resorption and osteoblastic formation of bone during cycles of remodeling; however this balance is lost with age. Trabecular bone loss is evident from the age of 30,2 and by the age of 50, up to 42% of total lifetime trabecular bone is lost.3 Contributing to this imbalance is a reduction in the level of bone formation during remodeling cycles, thought to occur as early as 20 years of age4 and is partly associated with reduced osteoblast differentiation and deposition of matrix components.5 Preventative intervention strategies are therefore paramount for maintaining skeletal homeostasis and, thus, life-long skeletal health. Epidemiological evidence suggests a positive association between fruit and vegetable intake and the accrual and maintenance of bone mass.6 Although the exact role(s) of fruit and vegetables remains unclear, much evidence from in vitro and in vivo studies suggests that flavonoids and related dietary phytochemicals could play an important role in supporting skeletal homeostasis via direct actions on osteoclasts and osteoblasts. The flavonols quercetin and kaempferol have gained considerable attention, with in vitro reports of their antiosteoclastogenic activity7−9 and the ability of both quercetin10,11 and kaempferol10−14 to promote osteoblast differentiation and function in several in vitro osteogenesis models. These observations are reinforced in vivo, where © XXXX American Chemical Society and American Society of Pharmacognosy

Received: June 15, 2015

A

DOI: 10.1021/acs.jnatprod.5b00075 J. Nat. Prod. XXXX, XXX, XXX−XXX

Journal of Natural Products

Article

Figure 1. Structure of compounds screened in the model of osteoblast differentiation. Compounds are numbered 1 to 21, and their respective substituents (R-groups) are highlighted in the table. Numbers within skeletal structures represent atom nomenclature.

Alongside flavonols, flavanones have attracted attention due to the positive effects of a hesperetin glycoside-enriched diet on bone health in OVX-mice,23 OVX-rats,24,25 and healthy young rats,25 with reports of suppressed osteoclast number and bone turnover and promotion of osteoblastic activity. Certain flavones26,27 and flavanols28 have also been seen to have the potential to influence bone health in vivo, adding to the weight of literature exclaiming flavonoids as possible therapeutic agents for the prevention of bone thinning. Although a compelling argument exists for certain flavonoids exerting bone-anabolic effects, the wide-ranging in vitro models of osteogenesis examined and concentrations tested (Table S1, Supporting Information) make identifying which flavonoids have the greatest effects and the structural basis for such activity difficult. Moreover, results using animal cells and transformed cell lines may be influenced by interspecies variation29 and variable phenotypic profiles,30 respectively. Thus, we investigated the influence of flavonoids and related dietary phytochemicals on the osteogenic differentiation of primary human bone-marrow-derived mesenchymal stem cells (hMSCs), the common precursor of the osteogenic lineage. We report the differential effects of a panel of 21 flavonoids and related phytochemicals, spanning the major flavonoid subclasses, on osteogenic differentiation within hMSC cultures

using alkaline phosphatase (ALP) activity. ALP activity is an established marker of early osteoblastogenesis, being expressed between the late osteoprogenitor and preosteoblast stages of commitment, increasing throughout the course of differentiation into mature osteoblasts, and subsequently declining throughout the mineralization period.31,32 As the protein demonstrating the greatest increase during early osteoblastogenesis,33 ALP activity was considered an excellent marker to screen for osteogenic activity. From these screening results, we have employed three-dimensional quantitative structure− activity relationship (3D QSAR) modeling with comparative molecular field analysis (CoMFA) to relate the observed pharmacological activity of these compounds to their chemical structures. Here we describe, for the first time, the effects of multiple flavonoids on the osteogenic differentiation of hMSCs and discuss the structural basis for their activity. Briefly, seven flavonols (tamarixetin, kaempferide, kaempferol, galangin, quercetin, fisetin, and isorhamnetin) and two flavanones (hesperetin and naringenin) stimulated osteogenic differentiation of hMSCs. CoMFA contour maps emphasized the importance of electronegativity, steric bulk, and a 2-C−3-C double bond at the flavonoid C-ring, as well as overall electropositivity and reduced steric bulk at the flavonoid B-ring. B



Article

Figure 2. Effects of dietary flavonoids and related phytochemicals on the osteogenic differentiation and proliferation of hMSCs. Passage three MSCs were subcultured in osteogenic differentiation (Os.D) medium and treated with 0−10 μM phytochemicals at days 0, 3, and 6. Cultures were assayed for ALP activity (A) and cell number (B) at day 9. Values are expressed as means of percentage of untreated control from three independent triplicate experiments from a single donor population of cells, with error bars representing SEM and asterisks representing significance compared to the vehicle control value (***p < 0.001; **p < 0.01; *p < 0.05). Phytochemical names corresponding to treatment numbers are displayed in Figure 1.

■

The order of magnitude was slightly different following 10 μM treatment (tamarixetin (1) > kaempferol (2) > galangin (3) > quercetin (4) > fisetin (5) > isorhamnetin (6)) with increases of between 3.01- and 1.32-fold over the VC; only tamarixetin (1), kaempferol (2), and galangin (3) induced a dosedependent response between 5 and 10 μM. In contrast to other flavonols tested, at 5 and 10 μM, myricetin (7) significantly reduced ALP activity by approximately 0.74- and 0.62-fold of the VC. No flavanone significantly increased ALP activity above the VC following 1 or 5 μM treatment, with only hesperetin (8) and naringenin (9) inducing a significant increase following 10 μM treatment (1.72- and 1.53-fold, respectively). Neither eriodictyol (10) nor homoeriodictyol (11) had a significant effect on ALP activity. Of the flavones tested in this study, diosmetin (12), apigenin (13), and flavone (14) had no effect on ALP activity at any concentration. Luteolin (15) induced a significant decrease in ALP activity following 10 μM treatment to approximately 0.59fold of the VC. No flavanol tested significantly increased ALP activity at any concentration. While (2R,3R)-epicatechin (16) and (2R,3S)-

RESULTS AND DISCUSSION Effect of Flavonoids on the Osteogenic Differentiation of hMSCs. Twenty-one flavonoids and related dietary compounds (Figure 1) were screened for effects on osteogenic differentiation, using ALP activity as a marker (Figure 2A). Initial experiments demonstrated that the highest increase in ALP activity occurred within the first 9 days of treatment with a number of test compounds and a positive control (Figure S1, Supporting Information). This, accompanied by a lower variability between experiments at day 9, meant that this time point was selected for subsequent screening of compounds for the structure−activity relationship analysis. No flavonol increased ALP activity compared to the vehicle control (VC) following 1 μM treatment; significant increases were observed following 5 and 10 μM treatments with all flavonols except myricetin (7). The 5 μM treatment with active flavonols increased ALP activity to approximate levels of between 2.37- and 1.52-fold compared to VC in the following order of magnitude: kaempferol (2) > tamarixetin (1) > quercetin (4) > fisetin (5) > galangin (3) > isorhamnetin (6). C



Article

Table 1. Effects of Phytochemical Treatment on the Viability of hMSCsa percentage viability day 1 # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

1 μM

VC 97 97 99 97 101 99 98 103 97 99 99 97 100 99 99 98 99 99 99 99 100

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

1 1 1 0 1 1 1 4 0 2 1 1 1 1 1 2 1 1 1 1 2

105 96 105 101 98 105 100 102 104 102 106 97 98 102 100 92 105 101 106 103 99

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

day 4 5 μM

6 3 4 5 4 4 2 3 5 4 5 3 2 3 3 6 4 2 4 2 9

102 104 105 100 98 107 94 95 105 98 107 89 102 103 96 90 104 98 105 101 101

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

10 μM 5 3 4 6 5 3 5 3 4 6 5 6 3 3 4 7 5 1 4 3 7

102 104 104 89 86 108 101 96 105 99 106 87 98 106 93 84 104 90 106 102 98

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

7 4 4 4 6 3* 5 1 4 4 4 9 3 4 2 6* 5 1* 4 4 6

1 μM

VC 98 99 97 100 98 97 98 100 100 99 97 98 98 97 99 98 97 97 97 97 99

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

day 7

1 1 2 2 3 2 1 0 2 1 2 0 1 2 0 2 2 2 2 2 1

97 97 94 98 96 93 97 98 101 103 95 99 101 93 99 98 94 93 93 92 100

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

5 μM 1 2 1 2 2 2 1 3 1 3 2 1 4 3 1 1 1 3 2 3 2

97 98 91 100 97 92 97 102 104 102 94 92 99 93 96 95 92 91 93 92 100

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

10 μM 2 1 2 3 3 1 1 3 2 2 2 5 4 3 1 2 2 4 1 3 2

93 98 91 101 95 91 101 103 106 102 98 90 98 92 94 94 93 89 95 92 100

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

3 2 2 3 5 3 1 2 1 1 1 6 2 2 0 3 1 6 1 4 2

1 μM

VC 98 96 99 99 100 99 98 101 100 99 99 99 99 99 100 101 99 99 99 99 98

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

1 0 0 2 1 0 1 1 2 2 0 2 2 0 1 2 0 0 0 0 3

94 100 99 103 98 98 97 97 100 101 100 98 96 100 97 96 97 101 97 98 96

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

5 μM 1 3 1 3 2 2 0 2 1 0 2 1 1 1 2 1 1 1 1 1 3

96 100 97 104 97 100 98 99 102 102 99 93 97 102 94 98 98 101 98 98 97

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

10 μM 2 4 1 4 3 3 1 2 3 1 2 3 3 0 1 1 1 1 1 2 3

93 97 97 101 96 97 98 98 103 101 102 93 94 98 96 98 98 102 100 101 97

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0 2 1 4 3 1 1 3 2 1 2 3 2 0 1 1 1 0 2 2 5

a Passage 3 MSCs were subcultured in osteogenic differentiation (Os.D) medium and treated with 0−10 μM polyphenols at days 0, 3, and 6. Cultures were assayed for cell viability after 1, 4, and 7 days (24 h after every treatment). Values are expressed as means of percentage of control ± SEM from three independent triplicate experiments, and asterisks represent significance compared to the vehicle control value (*p < 0.05). Phytochemical names corresponding to treatment numbers (#) are displayed in Figure 1.

catechin (17) had a negligible effect, (2R,3R)-epigallocatechin3-O-gallate (18) induced a significant decrease in ALP following 10 μM treatment to approximately 0.75-fold of the VC. Of the related compounds tested, neither phloretin (19), (2R,3R)-taxifolin (20), nor resveratrol (21) had an effect on ALP activity at any concentration tested. Effect of Flavonoids on the Proliferation and Viability of hMSCs. Effects of treatments on cell number (Figure 2A) and viability (Table 1) were also examined. Of the compounds tested, only kaempferol (2) and hesperetin (8) significantly increased cell number above the level of the VC. Five flavonols (tamarixetin (1), quercetin (4), fisetin (5), isorhamnetin (6), and myricetin (7)), four flavones (diosmetin (12), apigenin (13), flavone (14), and luteolin (15)), one flavanol ((2R,3R)epigallocatechin-3-O-gallate (18)), and the stilbene resveratrol (21) decreased cell number. With the exception of the flavanols (2R,3R)-epicatechin (16) and (2R,3R)-epigallocatechin-3-Ogallate (18), no treatment significantly reduced cell viability, suggesting antiproliferative activity of a number of flavonoids, most notably within the flavone subclass, with decreases in cell number as low as 0.43 ± 0.05-fold of the VC. QSAR Model Generation for the Relationship between Flavonoid Structure and ALP Activity. Using 3D QSAR modeling with CoMFA, models were developed that predict the effect of phytochemical treatment at 5 and 10 μM concentrations on ALP activity in hMSC cultures. The training set consisted of ALP activity data (expressed as log10 ALP fold increase over the vehicle control) for 18 compounds (refer to Figure 1 for training set structures). Kaempferol (2), eriodictyol (10), and EGCG (18) data were removed to establish a test set. Both low-energy (built) conformers and PDB-extracted

conformers (Table S3, Supporting Information) for each compound were aligned to flavone by overlapping atoms 1, 5−10, and 1′−6′. This allowed a satisfactory superimposition of all major areas of the compound, importantly, B- and C-rings of flavonoids, where the majority of structural variation in the training set occurred (Figure 3).

Figure 3. Alignment rule. Conformers were aligned by common core alignment to flavone, overlapping atoms 1, 5−10, and 1′−6′ (shaded, panel A). This delivered an alignment with appropriate superimposition of flavonoid A-, B-, and C-rings (see lateral view, panel B), as well as directional alignment of functional groups (oxygen atoms of hydroxyl and methoxy groups highlighted in black).

Statistical models were generated using partial least-squares (PLS) regression analysis. Linear regression plots between observed activity within the training set and predicted activity based on molecular descriptors had a good correlation at both 5 μM (r2 = 0.768; Figure 4A, Table 2) and 10 μM (r2 = 0.767; Figure 4B, Table 2), signifying that variances in molecular fields account for approximately 77% of the variances in training set activity. Table S2, Supporting Information, presents all CoMFA-generated values including residual values (mean D



Article

Figure 5. Test set compound structures. The structures and respective substituents (R-groups) of test compounds with known activity (2, 10, and 18) or unknown activity (T1, T2, and T3) are indicated. Numbers within skeletal structures represent atom nomenclature.

76% of the mean observed activity in cultures treated with 5 and 10 μM kaempferol (2), respectively. Models also overestimated the detrimental effect of EGCG (18), with predicted activity approximately 1.9- and 2.6-fold lower than actual activity in 5 and 10 μM models, respectively (Table 3).

Figure 4. Linear correlation between observed and predicted values generated from CoMFA for ALP activity at 5 μM (A) and 10 μM (B). Observed values (x-axis) are correlated with CoMFA-generated predicted values based on steric and electrostatic field variations. Predicted values are expressed as the mean of all individual conformer predictions, and numbers indicate compounds (as listed in Figure 1). Corresponding CoMFA-generated values are presented in Table S2, Supporting Information.

Table 3. Known Test Set Predicted Values and Observed ALP Activity Values 5 μM model

Table 2. Partial Least Squares Analysis for CoMFA Models d

5 μM model 10 μM model

r2a

q2b

SEc

compd

#

0.768 0.767

0.466 0.420

0.077 0.098

5 5

a

2 10b 18c

a

Correlation coefficient. bCross-validated correlation coefficient. Standard error for the prediction of log10(ALP fold change). d Number of components used in CoMFA model. c

obsd (%)

237* 98 84

g

10 μM model

e

pred (%) 185 110 44

d

resf (%) 52 −16 40

obsd (%)

268* 111 75*g

g

prede (%)

resf (%)

205 107 29

63 4 46

a

Kaempferol (2). bEriodictyol (10). cEGCG (18). dObserved. Predicted. fResidual. gp < 0.05 compared to control when tested experimentally.

e

observed value minus mean predicted value) and standard deviations of prediction between different conformers processed per compound. Both models had an optimum component number of 5 (Table 2). The predictive potential of the model was quantified using leave-one-out cross-validation. q2 values for 5 and 10 μM models were 0.466 and 0.420, with standard errors of prediction of 0.077 and 0.098, respectively (Table 2). Although these appear low, the q2 value is a conservative estimate, as its calculation involves removing data so true predictability can be underestimated.34 Therefore, models were further validated using a test set of molecules. Test Set of Molecules for Model Validation. The test set consisted of three compounds with known activity [“known compounds”: kaempferol (2), eriodictyol (10), and EGCG (18)], previously removed from the training set, and three with unknown activity [“blind compounds”: kaempferide (T1), pinocembrin (T2), and tricetin (T3)], which were predicted using the QSAR models and assayed in the model of osteoblast differentiation at a later date (see Figure 5 for structures). QSAR model predictions for compounds with known activity correctly identified kaempferol (2) as being active, eriodictyol (10) as having a negligible effect, and EGCG (18) as having a detrimental effect on ALP activity in hMSCs after 9 days in culture. Both 5 and 10 μM models are relatively conservative in their prediction of the level of ALP activity in response to kaempferol (2), with predicted values approximately 78% and

Of the blind compounds within the test set, kaempferide (T1) was predicted to increase ALP activity at 5 μM (180% of VC) and 10 μM (248% of VC), pinocembrin (T2) to have a slight detrimental effect on ALP activity at 5 μM (81% of VC) but negligible effects at 10 μM (96% of VC), and tricetin (T3) to decrease ALP activity at 5 and 10 μM (90% and 64% of VC, respectively). When tested experimentally, kaempferide (T1) increased ALP activity in hMSC cultures dose dependently with significant elevation following treatment with 1, 5, and 10 μM to 143 ± 6%, 237 ± 23%, and 281 ± 30% of the VC, respectively (Figure 6A). As with kaempferol (2), QSAR models underestimated activity, with predicted values 71% and 88% of the observed value in the 5 and 10 μM models, respectively (Table 5), but successfully identified an active compound. Pinocembrin (T2) had no significant effect on ALP activity at any concentration tested (Figure 6A). Although the exact values for pinocembrin (T2) predicted by QSAR models are lower than actual values for ALP activity (Table 5), the successful prediction of an inactive compound is highlighted. Finally, tricetin (T3), while having no significant effect on ALP activity at 1 or 5 μM, significantly decreased ALP activity following 10 μM treatment to a level of 79 ± 7% of the VC (Figure 6A). These predictions accurately described tricetin (T3) as a compound that has a detrimental effect on osteogenic differentiation at the higher concentration tested (Table 5). E



Article

Table 5. Blind Test Set Predicted Values and Observed ALP Activity Values 5 μM model

10 μM model

#

obsdd (%)

prede (%)

resf (%)

obsdd (%)

prede (%)

resf (%)

T1a T2b T3c

237*g 107 99

170 81 90

67 26 9

281*g 121 79*g

248 96 64

33 25 15

a

Kaempferide (T1). bPinocembrin (T2). cTricetin (T3). dObserved. Predicted. fResidual. gp < 0.05 compared to control when tested experimentally.

e

observed results follow these predictions, although estimates for active compounds in the test set are relatively conservative at both 5 and 10 μM (summarized in Figure 7A and C,

Figure 6. Effects of blind test compounds on the osteogenic differentiation and proliferation of hMSCs. Passage 3 MSCs were subcultured in osteogenic differentiation (Os.D) medium and treated with 0−10 μM kaempferide (T1), pinocembrin (T2), and tricetin (T3) at days 0, 3, and 6. Cultures were assayed for ALP activity (A) and cell number (B) at day 9. Values are expressed as means of percentage of control from three independent triplicate experiments from a single donor population of cells, with error bars representing SEM and asterisks representing significance compared to the vehicle control value (***p < 0.001; *p < 0.05).

Table 4. Effect of Blind Test Set on Viability of hMSCsa day

#

1

T1 T2 T3 T1 T2 T3 T1 T2 T3

1 μM

VC 102 102 102 100 100 100 98 98 98

± ± ± ± ± ± ± ± ±

1 1 1 1 1 1 0 0 0

112 109 108 95 94 99 97 96 97

± ± ± ± ± ± ± ± ±

3** 6 5 0 1 1 1 1 1

5 μM 112 110 106 95 95 95 98 97 94

± ± ± ± ± ± ± ± ±

3** 5 4 1 1 3 0 1 2

10 μM 111 111 104 95 96 95 99 98 94

± ± ± ± ± ± ± ± ±

3** 4 6 1 1 3 0 1 1

a

Figure 7. Correlation between observed and predicted activities in the test set. Observed and predicted values for the ALP activity of known and blind compounds (kaempferol (2), eriodictyol (10), EGCG (18), kaempferide (T1), pinocembrin (T2), tricetin (T3)) are presented as percentage of vehicle control in response to 5 μM (A) and 10 μM treatment (C). The respective linear correlations between observed and predicted values are represented for the 5 μM (B) and 10 μM (D) QSAR models.

Common to the majority of other flavonols tested, kaempferide (T1) did induce a slight decrease in cell number at 10 μM, although this was not significant. Where pinocembrin had no effect on cell number, tricetin (T3) decreased cell number at both 5 and 10 μM concentrations to levels of 0.68 ± 0.02- and 0.47-fold of the vehicle control, respectively (Figure 6B). No blind compound tested significantly decreased the viability of hMSC cultures (Table 4). In summary, QSAR models successfully predicted kaempferol (2) and kaempferide (T1) as being active, eriodictyol (10) and pinocembrin (T2) as having little effect, and EGCG (18) and tricetin (T3) as having little or a detrimental effect. The

respectively). The linear correlation between observed and predicted activity for known and blind predictions is relatively good for 5 μM (r2 = 0.895) and 10 μM (r2 = 0.995) models (Figure 7B,D), demonstrating the predictive capacity of QSAR models. Examination of Structure−Activity Relationships. CoMFA contour maps for 5 and 10 μM models were examined for regions where variations in steric and electrostatic molecular fields contribute to changes in training set compound activity. Maps for both models were essentially identical (inset in Figure 8Ai), also reflected by the similarity in r2 values (0.768, 5 μM; 0.767, 10 μM). Therefore, the following description of contours is considered applicable to both models. Figure 8 presents the contour maps for the most active training set conformer, tamarixetin (1), and the least active conformer, luteolin (15). Steric fields (Figure 8Ai,ii) suggest that the largest region where steric bulk is favorable (green) is at the C-3 and C-4 positions of the C-ring. Contributing to steric bulk here is the hydroxyl group at the third carbon (OH-

4

7

Passage 3 MSCs from were subcultured in osteogenic differentiation (Os.D) medium and treated with 0−10 μM polyphenols at days 0, 3, and 6. Cultures were assayed for cell viability after 1, 4, and 7 days (24 h after every treatment). Values are expressed as means of percentage of control ± SEM from three independent triplicate experiments, and asterisks represent significance compared to the vehicle control value (**p < 0.01). Phytochemical names corresponding to treatment numbers (#) are displayed in Figure 5.

F



Article

Indeed, a 2-C−3-C double bond at the C-ring appears crucial for flavonol activity, as exemplified by the comparison of quercetin (4) with taxifolin (20). While quercetin has a 2-C−3C double bond with associated planarity, taxifolin does not, resulting in flexibility and subsequent inactivity. An overlay of quercetin and taxifolin (purple) is displayed in Figure 8B, showing the unfavorable conformation of taxifolin, with OH-3 and oxo-4 groups positioned away from regions of steric and electronegative favorability. In the absence of an OH-3 group, however, it appears favorable to have a hydrogenated 2-C−3-C bond, as highlighted by the comparison of the flavanones hesperetin (8) and naringenin (9), both active at 10 μM, with the flavones diosmetin (12) and apigenin (13), both inactive at 10 μM. At the B-ring, there is a small steric-favorable region at the C4′ position that corresponds well with the methoxy group present in tamarixetin (1) (Figure 8Ai) but not in luteolin (15) (Figure 8Aii). The benefit of C-4′-methylation is supported by the 1.4-fold greater activity of tamarixetin (1) compared to quercetin (4) at 10 μM and the high activity of the test compound kaempferide (T1). There is also a small stericfavorable region at the C-3′ position, likely associated with the plane in which conformers are lying within contours, with the luteolin conformer’s OH-3′ being brought closer to the yellow unfavorable region (Figure 8Ai,ii). The steric-unfavorable contour at the 3′ position can be further associated with the detrimental effect of bulky substitution here, as illustrated by the 1.4-fold lower activity of isorhamnetin (6) (OCH3-3′) compared to quercetin (4) (OH-3′) at 10 μM. Also unfavorable is steric bulk at the 5′ position of the B-ring (Figure 8Ai,ii), where substitution here [e.g., myricetin (7)] was associated with a low level of activity. Also at the B-ring, there is a large blue contour covering the 3′, 4′, and 5′ positions, a region where positive charge is favorable for activity, whereas negative charge is unfavorable. This is highlighted by comparing the most active (Figure 8Aiii) and least active (Figure 8Aiv) compounds in the training set, where an OH-3′, OCH3-4′ configuration conveys better activity than OH-3′, OH-4′ substitution. A similar relationship occurs within flavonol and flavanone subclasses (i.e., tamarixetin (1) > quercetin (4); hesperetin (4) > eriodictyol (10)). Multiple hydroxylation at the B-ring was associated with the attenuation of activity, as exemplified by myricetin (7), EGCG (18), and tricetin (T3), where hydroxyl groups at 3′, 4′, and 5′ positions, conveying strong electronegativity, were unfavorable for activity. In addition to contours at the C and B rings, there are also steric-favorable and -unfavorable regions at the fifth and seventh positions of the A-ring (Figure 8Ai,ii). The majority of compounds within the training set were hydroxylated at these positions, so this is likely due to compound flexibility. This is illustrated by the comparison of quercetin (4) with its nonplanar analogue taxifolin (20) (Figure 8B); where the OH7 of quercetin is closer to the steric-favorable region, the OH-7 and OH-5 of taxifolin are closer to steric-unfavorable regions. CoMFA reveals two major molecular regions that contribute to osteogenic activity within the training set (Figure 8C). Perhaps the most important is at the C-ring, with OH-3 and oxo-4 groups conveying favorable steric bulk and electronegativity, as well as an unsaturated 2-C−3-C bond resulting in planarity. These characteristics are common to flavonols that exhibited the greatest activity within the training set. The second major region is at the B-ring, where steric bulk is

Figure 8. QSAR contour maps derived from CoMFA for ALP activity in response to flavonoid treatment. (A) Contour maps representative of the most active compound, tamarixetin (1a) (Ai and Aiii), and least active compound, luteolin (15b) (Aii and Aiv), in the 10 μM training set. Inset figure represents contours for the 5 μM training set. (B) Contour maps displaying the influence of a 2-C−3-C double bond on conformation. Green and yellow contours highlight steric fields (green: steric bulk is favorable; yellow: steric bulk is unfavorable). Blue and red contours highlight electrostatic fields (blue: positive charge is favorable; red: negative charge is favorable). Contours are calculated as the model Stdev*Coeff. (C) Summary of molecular regions, derived from CoMFA, important for the osteogenic activity of flavonoids: (I) steric bulk (shaded green) and electronegativity (shaded red) at the C3 and C-4 positions of the C-ring; (II) unsaturation of the 2-C−3-C of the C-ring (shaded gray) ensuring planarity; (III) an overall positive charge at the B-ring (shaded blue); (IV) steric bulk at the C-4′ position of the B-ring (shaded green); and (V) reduced steric bulk at the C-3′ and C-4′ positions of the B-ring (shaded yellow).

3), present in the flavonol tamarixetin (Figure 8Ai) but not in the flavone luteolin (Figure 8Aii). This region is also associated with a red contour, suggesting negative charge is favorable, supporting the importance of the OH-3 group (Figure 8Aiii,iv). The 3.1-, 2.8-, and 3.6-fold greater activity of the flavonols tamarixetin (1), kaempferol (2), and quercetin (4) compared to their respective flavone analogues diosmetin (12), apigenin (13), and luteolin (15) at 10 μM highlights the benefit of an OH-3. Also associated with these contours is the ketone group at the fourth position of the C-ring (oxo-4), although the only compounds without an oxo-4 group [catechin (17) and epicatechin (18)] are nonplanar due to 2-C−3-C bond hydrogenation, so a direct comparison cannot be made. G



Article

unfavorable at all positions except from the 4′ position, where bulky substitution may be favorable. Accompanied with this is an overall positive charge of the B-ring, suggesting that multiple hydroxylation is unfavorable for activity. Although the exact target(s) for flavonoids in our model of osteoblast differentiation is unknown, these variances in steric and electrostatic interactions are likely to govern binding affinity to target protein(s) and subsequently effect their biological activity.35 Pathways controlling osteoblast differentiation and bone formation are complex and highly regulated. Accompanying this is the apparent nonspecificity of flavonoids; flavonoids have been reported to bind to a variety of targets, including receptors, transcription factors, enzymes, and, importantly, a large number of protein kinases.36−40 The flavonol quercetin alone has been reported to inhibit the activity of a considerable number of kinases,38,39,41 many of which play a role in pathways controlling osteoblastogenesis. Moreover, inhibition of several of these kinases, including mitogen-activated protein kinase kinase 1 (MEK1),42 protein kinase C (PKC),43 and glycogen synthase kinase-3β (GSK3β), has been reported to promote osteoblast differentiation and/or bone formation.44 It is therefore likely that those flavonoids that have demonstrated activity in our model exert their effect via a number of different targets, culminating in the promotion of osteogenic differentiation. Although not the primary focus of this study, some intriguing structure−activity relationships have also been observed related to the effects of flavonoids on proliferation. Within the flavonol subclass, the absence of hydroxylation or hydroxymethlation at the B-ring was associated with negligible effects on proliferation, as was the case for galangin (3). Interestingly, however, OH-4′ substitution (kaempferol (2)) increased proliferation, while this effect was abolished by methylation of the OH-4′ (kaempferide (T1)). Substitution at the C-4′ position, whether by a hydroxyl or methoxy group, was associated with antiproliferative activity, with tamarixetin (1), quercetin (4), fisetin (5), isorhamnetin (6), and myricetin (7) all decreasing cell number while having no effect on cell viability. This suggests that substitution at the B-ring could play a considerable role in the antiproliferative activity of flavonols. This observation was not as clear in the case of flavones, however, where each flavone had antiproliferative activity. With flavonoids being the basis for drug design of antiproliferative agents in cancer therapeutics,45 these results may contribute to understanding the molecular properties of flavonoids that convey effects on proliferation. Effect of Kaempferide (T1) on Matrix Mineralization. In order to further support the capacity of the QSAR model to describe and predict osteogenic compounds, kaempferide (T1) was selected to examine its capacity to support matrix mineralization, indicating augmentation of mature bone formation. As the compound with the highest predicted level of osteogenic activity in the test set of compounds, kaempferide (T1) conformed well to the contour maps derived from CoMFA. While control (VC) cultures exhibited very weak levels of matrix mineralization as assessed by Alizarin Red S staining of calcium deposition, treatment with 10 μM kaempferide (T1) resulted in a high level of calcium deposition across the cell monolayer (Figure 9A). Upon elution and quantification of Alizarin Red S, kaempferide (T1) treatment was demonstrated to significantly increase matrix mineralization to 3.17-fold of the VC. This considerable increase in mineralization supports the potential of the model to predict

Figure 9. Effect of kaempferide (T1) on matrix mineralization. Passage 3 MSCs were subcultured in osteogenic differentiation (Os.D) medium supplemented with 10 mM β-glycerophosphate and treated with 10 μM T1 at days 0, 3, 6, 9, and 12. Cultures were maintained for 20 days before staining with Alizarin Red S. Photographs of stained monolayers were captured (A) before bound Alizarin Red S was eluted and quantified. Values are expressed as means of percentage of untreated control from three independent triplicate experiments from a single donor population of cells, with error bars representing SEM and asterisks representing significance compared to the vehicle control value (*p < 0.05).

the osteogenic activity of phytochemicals previously undescribed.

■

CONCLUSIONS Seven flavonols [tamarixetin (1), kaempferol (2), kaempferide (T1), galangin (3), quercetin (4), fisetin (5), and isorhamnetin (6)] and two flavanones [hesperetin (8) and naringenin (9)] were found to significantly augment ALP activity in hMSC cultures here for the first time. All other compounds belonging to flavones, flavanols, dihydroflavonols, dihydrochalcones, or stilbenes did not significantly increase ALP activity above the vehicle control. Although many of the compounds tested in our model have previously been investigated for their effect on osteoblastogenesis, outcomes have been highly variable (summarized in Table S1, Supporting Information), demonstrating the requirement for comprehensive screening within a single model. Through executing this, valuable insights have been realized with respect to the molecular properties of flavonoids that convey osteogenic activity in hMSC cultures. The consistent effects of flavonols in our model corroborate the promising observations of kaempferol (2) and quercetin (4) increasing osteogenic differentiation in vitro.10−14 This, accompanied by the osteogenic effects of the two flavonols in in vivo models,10,13,15−21 highlights an excellent potential for dietary flavonoids belonging to this subclass to promote skeletal health, either by maximizing bone mass accrual during growth or by modulating skeletal homeostasis throughout life, thus reducing the lifetime risk of fragility fractures.2 In the context of dietary achievability, we have demonstrated that the majority of active compounds in our model, predominantly flavonols, exhibit osteogenic activity at concentrations in the range 5 to 10 μM. Concentrations of the flavonol quercetin can reach between 5 and 8 μM in human plasma following ingestion, with levels achieved depending on the nature of the glycoside,46,47 and long half-lives of ingested flavonoids suggest that repeated intake might lead to bioaccumulation. However, it must be recognized that achieving these levels through dietary intake would be challenging, as flavonoids are rapidly metabolized following H



Article

0.1% (v/v) DMSO, and all assays also included a vehicle control [0.1% (v/v) DMSO only], a positive control (50 nM calcitriol), and an untreated control. All tests within individual experiments were performed in triplicate. Treated cultures were maintained for 9 days with media changes and reapplication of compounds every 3 days. Determination of Osteogenic Differentiation (ALP Activity), Cell Number, and Viability. Cultures were assessed for ALP activity and cell number on day 9, a time point in our model found to be optimal for minimal variability, and for cell viability on days 1, 4, and 7 (24 h after every treatment). ALP activity was determined using the para-nitrophenyl phosphate-based colorimetric method as described previously,55 and data were normalized to time and cell number (μM pNP/min/20 000 cells). Cell number was determined using the methylene blue-based proliferation assay as described previously,56 and cell viability was determined using the Alamar blue cytotoxicity assay as described previously.57 All data were expressed as the percentage of the untreated control value. Quantitative Structure−Activity Relationship Study. Threedimensional QSAR models were developed using SYBYL-X (v1.3) molecular modeling software (Tripos). CoMFA descriptors were selected to correlate steric and electrostatic field energy variations with ALP activity in response to 5 and 10 μM treatments. The training set of compounds consisted of 18 phytochemicals with their associated biological activity, and the predictive capacity of models was validated using a test set of six compounds. Determination of Matrix Mineralization. Cultures were maintained for 20 days in Os.D medium supplemented with 10 mM β-glycerophosphate with phytochemical treatment on days 0, 3, 6, 9, and 12. Matrix mineralization (calcium deposition) was assessed by Alizarin Red S staining, which was quantified by elution and colorimetric determination as described previously.58 Statistical Analyses. Statistical analysis was carried out using GraphPad Prism (v4.00). Data are expressed as means of percentage of the untreated control from three independent triplicate experiments ± standard error of the mean (SEM). Statistical comparisons were made using one-way analysis of variance (ANOVA; ALP activity and cell number) or two-way ANOVA (cell viability) followed by Bonferroni post hoc tests. p-Values < 0.05 between treatments and the vehicle control were considered significant (*p < 0.05; **p < 0.01; ***p < 001). For 3D QSAR, statistical analyses were generated using SYBYLX (v1.3) (Tripos), and regression models produced using PLS analysis with leave-one-out cross-validation.

intake, resulting in glucuronide, sulfoglucuronide, and sulfate conjugates.48 While this may compromise the bioactivity of the aglycone, it has been argued that such hydrophilic conjugation could facilitate the circulatory transport of active compounds that are released from their conjugates at target tissues through the action of enzymes such as β-glucuronidase in the vasculature.49−51 While debate remains over the dietary achievable levels of flavonoids in humans, our results accompanied by human bioavailability studies suggest that active flavonoids at the concentrations tested could contribute to the dietary prevention of bone deterioration. We have also demonstrated the capacity for QSAR models to predict the osteogenic activity of compounds, detailing their potential for the prediction of other flavonoids and related compounds with unknown activity and also for the lead development of novel bone-anabolic treatments. Natural products have been the source of lead compounds for the development of drugs targeted at a number of conditions.52 Gaining an understanding of the key molecular regions of flavonoids that convey pro-osteogenic activity forms a basis to make analogues with improved pharmacological and/or pharmacokinetic properties. The predominant treatments for osteoporosis are the antiresorptive bisphosphonates, and although in recent times new bone-anabolic treatments have been developed, these are associated with numerous sideeffects,53 highlighting the continued need for better anabolic treatments for bone-metabolic disorders.

■

EXPERIMENTAL SECTION

Full methods are available in the Supporting Information. Human Mesenchymal Stem Cell Culture. Frozen vials of fully characterized passage 1 hMSCs from the bone marrow of a 27-yearold, healthy male donor were obtained from the Adult Mescenchymal Stem Cell Resource at Texas A&M Health Science Center College of Medicine, Institute for Regenerative Medicine (NIH/NCRR grant 5P40RR017447-03). Cells supplied were immunophenotyped (CD34, CD36, CD45, and CD117 negative expression; CD105, CD29, CD49c, CD147, and CD59 positive expression), and their ability to form colony forming units (CFUs) and their osteo- and adipogenic differentiation potential was validated. After 24 h of recovery, hMSCs were expanded at low density (50 cells/cm2) in α-minimum essential medium containing 2 mM GlutaMAX supplemented with 12% (v/v) fetal bovine serum (FBS) and 1% (v/v) penicillin G (10 000 units/ mL)/streptomycin sulfate (10 000 μg/mL) (Life Technologies, UK) in a humidified incubator at 37 °C with 5% CO2, with medium changes every 3−4 days until 80% confluency was reached. The CFU capacity of expanded cultures was assessed using the crystal violet staining method as previously described.54 Cultures with a CFU capacity of less than 40% were not used. To assess the impact of test compounds on osteoblast differentiation, proliferation, cell viability, and mineralization, passage 3 cells were cultured at high density (10 000 cells/cm2) in osteogenic differentiation (Os.D) medium (lowglucose Dulbecco’s modified Eagle medium containing 3.97 mM GlutaMAX supplemented with 12% (v/v) FBS and 1% (v/v) penicillin G (10 000 units/mL)/streptomycin sulfate (10 000 μg/mL) (Life Technologies, UK), with 10 nM dexamethasone and 50 μM L-ascorbic acid-2-phosphate (Sigma-Aldrich, UK)). For mineralization cultures, Os.D medium was supplemented with 10 mM β-glycerophosphate (Sigma-Aldrich, UK). Cells were allowed to adhere overnight in a humidified incubator at 37 °C with 5% CO2 prior to treatment with test compounds or controls. Phytochemicals. Of the 24 purchased phytochemicals examined in this study, 22 were determined by HPLC to be ≥95% pure and two were ≥90% pure. All solubilized phytochemicals were serially diluted in Os.D medium to achieve test compound concentrations of 10, 5, and 1 μM. Treated wells were adjusted to have a final concentration of

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jnatprod.5b00075. (i) Summary of the osteogenic effect of flavonoids from previous in vitro studies; (ii) effects of dietary flavonoids and related phytochemicals on the proliferation of hMSCs; (iii) effects of dietary flavonoids and related phytochemicals on cell viability in hMSC cultures, (iv) CoMFA-generated values for ALP in QSAR models; (v) known test set predicted and observed ALP activity values; (vi) blind test set predicted values for ALP activity; (vii) training set conformer predictions; (viii) Supporting Information methods (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Tel: +44 (0) 118 378 7032. E-mail: [email protected]. Notes

The authors declare no competing financial interest. I



■

Article

(25) Horcajada, M. N.; Habauzit, V.; Trzeciakiewicz, A.; Morand, C.; Gil-Izquierdo, A.; Mardon, J.; Lebecque, P.; Davicco, M. J.; Chee, W. S. S.; Coxam, V.; Offord, E. J. Appl. Physiol. 2008, 104, 648−654. (26) Park, J. A.; Ha, S. K.; Kang, T. H.; Oh, M. S.; Cho, M. H.; Lee, S. Y.; Park, J. H.; Kim, S. Y. Life Sci. 2008, 82, 1217−1223. (27) Kim, T.-H.; Jung, J. W.; Ha, B. G.; Hong, J. M.; Park, E. K.; Kim, H.-J.; Kim, S.-Y. J. Nutr. Biochem. 2011, 22, 8−15. (28) Shen, C. L.; Yeh, J. K.; Cao, J. J.; Wang, J. S. Nutr. Res. (N. Y., NY, U. S.) 2009, 29, 437−456. (29) Czekanska, E. M.; Stoddart, M. J.; Richards, R. G.; Hayes, J. S. Eur. Cells Mater. 2012, 24, 1−17. (30) Pautke, C.; Schieker, M.; Tischer, T.; Kolk, A.; Neth, P.; Mutschler, W.; Milz, S. Anticancer Res. 2004, 24, 3743−3748. (31) Aubin, J. E.; Liu, F.; Malaval, L.; Gupta, A. K. Bone 1995, 17, S77−S83. (32) Miron, R. J.; Zhang, Y. F. J. Dent. Res. 2012, 91, 736−744. (33) Foster, L. J.; Zeemann, P. A.; Li, C.; Mann, M.; Jensen, O. N.; Kassem, M. Stem Cells 2005, 23, 1367−1377. (34) Tripos. Tripos Bookshelf 1.2; Tripos International: Missouri, USA, 2010. (35) Kubinyi, H. 3D QSAR in Drug Design: Theory, Methods and Applications; Kluwer/Escom: Dordrecht, The Netherlands, 1993. (36) Le Bourvellec, C.; Renard, C. M. G. C. Crit. Rev. Food Sci. Nutr. 2012, 52, 213−248. (37) Williams, R. J.; Spencer, J. P. E.; Rice-Evans, C. Free Radical Biol. Med. 2004, 36, 838−849. (38) Davies, S. P.; Reddy, H.; Caivano, M.; Cohen, P. Biochem. J. 2000, 351, 95−105. (39) Hou, D.-X.; Kumamoto, T. Antioxid. Redox Signaling 2010, 13, 691−719. (40) Virgili, F.; Marino, M. Free Radical Biol. Med. 2008, 45, 1205− 1216. (41) Middleton, E.; Kandaswami, C.; Theoharides, T. C. Pharmacol. Rev. 2000, 52, 673−751. (42) Ndong, J. D.; Stevens, D. M.; Vignaux, G.; Uppuganti, S.; Perrien, D. S.; Yang, X. L.; Nyman, J. S.; Harth, E.; Elefteriou, F. J. Bone Miner. Res. 2015, 30, 1118−1118. (43) Jeong, H. M.; Jin, Y. H.; Choi, Y. H.; Yum, J.; Choi, J. K.; Yeo, C. Y.; Lee, K. Y. Biochim. Biophys. Acta, Mol. Cell Res. 2012, 1823, 1225−1232. (44) Gilmour, P. S.; O’Shea, P. J.; Fagura, M.; Pilling, J. E.; Sanganee, H.; Wada, H.; Courtney, P. F.; Kavanagh, S.; Hall, P. A.; Escott, K. J. Toxicol. Appl. Pharmacol. 2013, 272, 399−407. (45) Ragab, F. A.; Yahya, T. A. A.; El-Naa, M. M.; Arafa, R. K. Eur. J. Med. Chem. 2014, 82, 506−520. (46) Graefe, E. U.; Wittig, J.; Mueller, S.; Riethling, A. K.; Uehleke, B.; Drewelow, B.; Pforte, H.; Jacobasch, G.; Derendorf, H.; Veit, M. J. Clin. Pharmacol. 2001, 41, 492−499. (47) Olthof, M. R.; Hollman, P. C. H.; Vree, T. B.; Katan, M. B. J. Nutr. 2000, 130, 1200−1203. (48) Lesser, S.; Wolffram, S. Curr. Top. Nutraceut. R. 2006, 4, 239− 256. (49) Terao, J.; Murota, K.; Kawai, Y. Food Funct. 2011, 2, 11−17. (50) Menendez, C.; Duenas, M.; Galindo, P.; Gonzalez-Manzano, S.; Jimenez, R.; Moreno, L.; Zarzuelo, M. J.; Rodriguez-Gomez, I.; Duarte, J.; Santos-Buelga, C.; Perez-Vizcaino, F. Mol. Nutr. Food Res. 2011, 55, 1780−90. (51) Galindo, P.; Rodriguez-Gomez, I.; Gonzalez-Manzano, S.; Duenas, M.; Jimenez, R.; Menendez, C.; Vargas, F.; Tamargo, J.; Santos-Buelga, C.; Perez-Vizcaino, F.; Duarte, J. PLoS One 2012, 7, e32673. (52) Harvey, A. L. Drug Discovery Today 2008, 13, 894−901. (53) Das, S.; Crockett, J. C. Drug Des., Dev. Ther. 2013, 7, 435−448. (54) DiGirolamo, C. M.; Stokes, D.; Colter, D.; Phinney, D. G.; Class, R.; Prockop, D. J. Br. J. Haematol. 1999, 107, 275−281. (55) Sabokbar, A.; Millett, P. J.; Myer, B.; Rushton, N. Bone Miner. 1994, 27, 57−67. (56) Oliver, M. H.; Harrison, N. K.; Bishop, J. E.; Cole, P. J.; Laurent, G. J. J. Cell Sci. 1989, 92, 513−518.

ACKNOWLEDGMENTS This work was jointly funded by the Biotechnology and Biological Sciences Research Council (BB/G530184/1) and GlaxoSmithKline. hMSCs for the study were provided by Prof. D. J. Prockop (Texas A&M Health Science Center College of Medicine Institute for Regenerative Medicine at Scott & White Hospital; NIH/NCRR grant 5P40RR017447-03).

■

REFERENCES

(1) Fernandez-Tresguerres-Hernandez-Gil, I.; Alobera-Gracia, M. A.; del-Canto-Pingarron, M.; Blanco-Jerez, L. Med. Oral Patol. Oral Cir. Bucal. 2006, 11, E151−7. (2) Khosla, S. J. Gerontol., Ser. A 2013, 68, 1226−1235. (3) Riggs, B. L.; Melton, L. J.; Robb, R. A.; Camp, J. J.; Atkinson, E. J.; Peterson, J. M.; Rouleau, P. A.; McCollough, C. H.; Bouxsein, M. L.; Khosla, S. J. Bone Miner. Res. 2004, 19, 1945−1954. (4) Raisz, L. G. Bone 2007, 40, S1−S4. (5) Nishida, S.; Endo, N.; Yamagiwa, H.; Tanizawa, T.; Takahashi, H. E. J. Bone Miner. Metab. 1999, 17, 171−177. (6) Prynne, C. J.; Mishra, G. D.; O’Connell, M. A.; Muniz, G.; Laskey, M. A.; Yan, L. Y.; Prentice, A.; Ginty, F. Am. J. Clin. Nutr. 2006, 83, 1420−1428. (7) Wattel, A.; Kamel, S.; Mentaverri, R.; Lorget, F.; Prouillet, C.; Petit, J. P.; Fardelonne, P.; Brazier, M. Biochem. Pharmacol. 2003, 65, 35−42. (8) Wattel, A.; Kamel, S.; Prouillet, C.; Petit, J. P.; Lorget, F.; Offord, E.; Brazier, M. J. Cell. Biochem. 2004, 92, 285−295. (9) Pang, J. L.; Ricupero, D. A.; Huang, S.; Fatma, N.; Singh, D. P.; Romero, J. R.; Chattopadhyay, N. Biochem. Pharmacol. 2006, 71, 818− 826. (10) Kim, Y. J.; Bae, Y. C.; Suh, K. T.; Jung, J. S. Biochem. Pharmacol. 2006, 72, 1268−1278. (11) Prouillet, C.; Maziere, J. C.; Maziere, C.; Wattel, A.; Brazier, M.; Kamel, S. Biochem. Pharmacol. 2004, 67, 1307−1313. (12) Byun, M. R.; Jeong, H.; Bae, S. J.; Kim, A. R.; Hwang, E. S.; Hong, J.-H. Bone 2012, 50, 364−372. (13) Trivedi, R.; Kumara, S.; Kumar, A.; Siddiqui, J. A.; Swarnkar, G.; Gupta, V.; Kendurker, A.; Dwivedi, A. K.; Romero, J. R.; Chattopadhyay, N. Mol. Cell. Endocrinol. 2008, 289, 85−93. (14) Yang, L.; Chen, Q.; Wang, F.; Zhang, G. J. Ethnopharmacol. 2011, 135, 553−560. (15) Horcajada-Molteni, M. N.; Crespy, V.; Coxam, V.; Davicco, M. J.; Remesy, C.; Barlet, J. P. J. Bone Miner. Res. 2000, 15, 2251−2258. (16) Tsuji, M.; Yamamoto, H.; Sato, T.; Mizuha, Y.; Kawai, Y.; Taketani, Y.; Kato, S.; Terao, J.; Inakuma, T.; Takeda, E. J. Bone Miner. Metab. 2009, 27, 673−681. (17) Wong, R. W. K.; Rabie, A. B. M. J. Orthop. Res. 2008, 26, 1061− 1066. (18) Yang, L.; Takai, H.; Utsunomiya, T.; Li, X.; Li, Z.; Wang, Z.; Wang, S.; Sasaki, Y.; Yamamoto, H.; Ogata, Y. J. Cell. Biochem. 2010, 110, 1342−1355. (19) Lucinda, L. M. F.; de Oliveira, T. T.; Salvador, P. A.; Peters, V. M.; Reis, J. E. P.; Guerra, M. O. Phytother. Res. 2010, 24, 264−267. (20) Lucinda, L. M. F.; Vieira, B. J.; Oliveira, T. T.; Sa, R. C. S.; Peters, V. M.; Reis, J. E. P.; Guerra, M. O. Fitoterapia 2010, 81, 982− 987. (21) Trivedi, R.; Kumar, A.; Gupta, V.; Kumar, S.; Nagar, G. K.; Romero, J. R.; Dwivedi, A. K.; Chattopadhyay, N. Mol. Cell. Endocrinol. 2009, 302, 86−91. (22) Harborne, J. B.; Baxter, H. The Handbook of Natural Flavonoids; John Wiley & Sons Ltd.: Chichester, 1999; Vol. 1. (23) Chiba, H.; Uehara, M.; Wu, J.; Wang, X. X.; Masuyama, R.; Suzuki, K.; Kanazawa, K.; Ishimi, Y. J. Nutr. 2003, 133, 1892−1897. (24) Habauzit, V.; Nielsen, I. L.; Gil-Izquierdo, A.; Trzeciakiewicz, A.; Morand, C.; Chee, W.; Barron, D.; Lebecque, P.; Davicco, M. J.; Williamson, G.; Offord, E.; Coxam, V.; Horcajada, M. N. Br. J. Nutr. 2009, 102, 976−984. J



Article

(57) O’Brien, J.; Wilson, I.; Orton, T.; Pognan, F. Eur. J. Biochem. 2000, 267, 5421−5426. (58) Gregory, C. A.; Gunn, W. G.; Peister, A.; Prockop, D. J. Anal. Biochem. 2004, 329, 77−84.

K


Defining Key Structural Determinants for the Pro ... - ACS Publications

Recommend Documents