In this study, we determined the possible estrogenic/antiestrogenic properties of 10H2DA, 3,10DDA and SA, isolated from RJ and identified by spectroscopic methods 
. In choosing the concentrations we considered 1) the commonly used RJ dietary supplementation (1–3 g daily), 2) the concentration of 10H2DA and the concentration of sebacic acid in RJ (3–6% and 0.5% respectively) 
, 3) the concentration of 10H2DA, sebacic acid and 3,10 DDA as well as 10HDA acid in marketed RJ samples in Greece (40–50%, 5%, 4% and 20% respectively), 4) the human blood volume and bioavailability. Based on the above information, we decided to examine the biological effects of FAs in a concentration range of 10−10
M, which are physiologically achievable concentrations.
Using a ChIP assay in MCF-7 breast cancer cells, which are stably transfected with an inducible version of ERβ and express endogenous ERα, we examined the ligand-dependent recruitment of ERα and ERβ to chromatin. None of the tested FAs could modulate ERα recruitment to the pS2 promoter, whilst they increased ERβ recruitment to this promoter. All FAs inhibited the effect of E2
on ERα and ERβ recruitment. Consistent with the effects on receptor recruitment to DNA, experiments revealed that in the presence of ERβ, FAs could decrease pS2 mRNA levels, when added alone, and that they decreased E2
's effect in the presence and absence of ERβ. However, since in this cell system endogenous ERα is always present, effects on pS2 expression cannot easily be determined for ERβ alone. We further assessed the effects of FAs on ERα alone and ERβ alone in HeLa cells. This cell line, in contrast to MCF-7 cells, lacks endogenous ER. In HeLa cells, we demonstrated that all FAs, when assayed alone, were weak enhancers of ERα-mediated activity, while they antagonized ERβ-mediated effects. In the presence of E2
they antagonized the E2
-mediated effects via ERα and ERβ. The well characterized selective estrogen receptor modulator (SERM) 4OH-TMX also exhibited agonistic effects on ERα-mediated activity, while it was a complete antagonist of ERβ-mediated action. This is in agreement with a previous study reporting that 4OH-TMX induced ERE-mediated reporter gene activity in a stably transformed ERα expressing cell line, but exhibited pure antagonism in the corresponding ERβ expressing system 
Recruitment of co-factors is an essential component of ER signaling. The best defined structure-function of a co-regulator interaction is with co-activators that interact through a conserved LxxLL motif, termed an NR box. Interestingly, in MCF-7 cells we show that the recruitment of the EAB1 co-activator peptide upon E2 binding is reduced when FAs are present. This suggests that FAs are preventing proper ER activity, possibly by inducing a conformational response at the co-activator binding site, leading to masking of the co-activator site.
In the ERE-driven luciferase reporter gene assay in MCF-7 cells, all 3 FAs inhibited the E2
-mediated increase in luciferase activity, suggesting an ER-mediated effect and a common signal transduction pathway for E2
and FAs at the level of ERE-containing promoters. Additionally, all 3 FAs showed a trend towards increasing the ERE-driven luciferase activity when tested alone. This is consistent with results from Suzuki et al. showing that 10H2DA increased the ERE-driven luciferase activity in MCF-7 cells at the same concentration range. However, co-incubation of FAs with E2 was not investigated in their study 
. In previous reports 
fresh RJ displays agonistic activity in the ERE-driven luciferase reporter gene assay in MCF-7 cells similar to that observed for E2
whereas the isolated FAs in our study show little agonist activity and possess antagonistic activity. RJ contains multiple FA components 
and data indicate that 10H2DA, sebacic acid and 3,10 DDA (investigated in this study) may not be the only FA determinants that predict estrogen/antiestrogen activity in RJ 
. Additionally, RJ may exhibit biological effects determined by synergistic and/or antagonistic interactions between its constituents thus showing different biological effects than the biological activity of its isolated components.
The specificity of FAs with regard to steroid receptor activation was explored by assaying the effects of FAs in MCF-7 cells on GRE-mediated transactivation. The FAs did not alter the basal nor the Dex-induced GRE-mediated transcriptional activity, indicating that the inhibition by FAs has specificity with respect to modulation of NR-mediated functions. In line with our findings, Thurmond et al. 
proposed that medium chain FAs (hexanoate) at high concentrations (mM range) interacted with ERs to inhibit ligand stimulated transcription, while there was no effect on GR-mediated activity. Previous reports have shown that short chain FAs (valproic acid or butyrate and methoxyacetic acid) may act as deacetylase inhibitors at high concentrations (mM range) resulting in the induction of transcriptional silencing of ERα expression, which would imply that they are antiestrogenic in MCF-7 cells 
. The antiestrogen effects of the above short chain FAs are considered an effect that may be due to their inherent HDAC inhibitory activities, since they have all been shown to reduce endogenous ERα expression and have been characterized as HDAC inhibitors. Interestingly, a recent report showed that methoxyacetic acid (MAA at mM concentrations) modulates ERα and ERβ-mediated signaling, lowers endogenous ERα expression and antagonizes E2
-stimulated expression of ERα target genes, yet it does not compete with E2
for binding to ERα 
. However, in our study, FAs (at µΜ concentrations) did not affect ERα mRNA or protein levels.
We have explored possible mechanism(s) for the effects of FAs on ER signaling by molecular modeling. As mentioned above, it is possible that the FAs compete with the LXXLL- containing co-activator for the activation function domain 2 (AF2) binding site of the receptor. Of note, docking experiments showed significant favorable interaction energy between the FAs and ERs. However, similar interaction energies were also observed for other locations on the protein's surface, distant from the co-activator binding site. Among the locations showing substantially more favorable intermolecular interactions (−211 kcal/mol) is a region including the loop around Tyr459. This loop is part of the subunit interface in the dimeric ER. Hence, binding of FAs may interfere with the dimerization of ERs and in this way influence co-activator binding ().
FAs may bind to the ligand pocket, thus competing with E2
. The computational fitting showed very good compatibility of the ligand pocket for all three FAs (). Although the calculated interaction energies between ligands and receptor are only indirectly related to binding affinities, they do indicate that, similarly to E2
, the three FAs interact favorably with the ER when they are in the ligand pocket. However, our competition binding study did not show any binding of SA and 3,10DDA and binding only at extreme concentrations (10−4
M) of 10H2DA, indicating that an interaction with ERs is not mediated via the ligand binding pocket. In agreement, Suzuki et al (2007) showed that 10H2DA had little effect (about 20% inhibition) upon the ability of E2
to bind to ERα and 50% inhibition of E2
to bind to ERβ at a concentration of approximately 100 µΜ 
In line with our findings, a recent study on 3,3′-diindolylmethane, a selective activator of ERβ that does not bind to ERβ, proposes a possible mechanism of activation through recruitment of co-activators (i.e. SRC-2) 
. Moreover, it has been shown that the methoxyacetic study which modulates ERα signaling yet does not bind to ERα 
. Of note, recent findings indicate that ligands, without binding affinity to ERα, activate GPR30 signaling and may act synergistically or may antagonize ERα-mediated gene expression 
. Future studies should address the potential of FAs to activate GPR30 signaling or phosphorylation pathways in cooperation with ERs.
On the basis of the findings by Narita et al. 
demonstrating that RJ stimulates bone formation, we used the osteoblastic cell line KS483 followed by the Alizarin Red-S staining as a model system to study the effect of FAs on the mineralization process 
, which is known to be an estrogen induced effect. The murine KS483 cell line is a mesenchymal precursor cell line, which differentiates into mature mineralizing osteoblasts during a three-week culture period, when cultured under osteogenesis inducing conditions. This differentiation process can be divided in a proliferation, matrix formation, matrix maturation and finally a mineralization phase, according to the model of Stein and Lian 
. Thus, the defining characteristic of the mature osteoblast is its ability to produce a mineralized bone matrix. Moreover, KS483 cell model is among few osteoblastic culture systems that can produce discrete, three-dimensionally organized mineralized matrices which are recognizably bone like. These bone nodules consist of woven bone matrix covered by cuboidal osteoblastic cells and containing osteocyte-like cells embedded in the matrix. Characterization of mineralized bone nodules has demonstrated that the processes of nodule formation, matrix deposition and subsequent mineralization follow a well ordered, temporally defined pattern which appears analogous to bone formation and mineralization in vivo.
Low concentrations of SA or 10H2DA significantly induced mineralization, which was suppressed by the addition of ICI182780, indicating an ER-mediated effect. As expected, the presence of E2
significantly stimulated the mineralization of osteoblasts 
. Our results imply that 10H2DA and SA may be the RJ components that stimulate osteoblasts. None of the FAs stimulated or inhibited cell viability/proliferation of endometrial cancer (Ishikawa) or breast cancer (MFC-7) cells (Fig. S5
). The antiestrogenic effect of FAs in breast cancer cells, their favorable effect on osteoblasts and the lack of effect on endometrial cell viability suggest that FAs may be potential natural SERMs.
RJ is used extensively in commercial nutritional supplements, medical products, and cosmetics in many countries, while SA, one of its major components, is widely employed in medical practice, e.g. parenteral nutrition, orthopedic applications, drug delivery systems, vaccine development 
. This honey bee-excreted biological fluid possesses estrogen-like activity, yet the compounds mediating its estrogenic effects are largely unknown. The present report investigated the effects of RJ-derived FAs, namely 10-hydroxy-2-decenoic, 3,10-dihydroxydecanoic and sebacic acid, on estrogen signaling () and suggests that these RJ-derived medium chain fatty acids, structurally entirely different from E2
, mediate estrogen signaling, at least in part, by modulating the recruitment of ERα, ERβ and co activators to target genes ().
Flow chart of assays and summary of findings.