Design, Synthesis and Biochemical Evaluation of Novel Selective Estrogen Receptor Ligand Conjugates Incorporating an Endoxifen-Combretastatin Hybrid Scaffold

Nuclear-receptors are often overexpressed in tumours and can thereby be used as targets when designing novel selective chemotherapeutic agents. To date, many conjugates incorporating an estrogen receptor (ER) ligand have been synthesised in order to direct chemical agents to tissue sites containing ERs. A series of ER ligand conjugates were synthesised incorporating an antagonistic ER ligand scaffold based on endoxifen, covalently-bound via an amide linkage to a variety of combretastatin-based analogues, which may act as antimitotic agents. These novel endoxifen-combretastatin hybrid scaffold analogues were biochemically evaluated in order to determine their antiproliferative and cytotoxicity effects in both the ER-positive MCF-7 and the ER-negative MDA-MB-231 human breast cancer cell lines. ER competitive binding assays were carried out to assess the binding affinity of the lead conjugate 28 towards both the ERα and ERβ isoforms. In results from the NCI 60-cell line screen, the lead conjugate 28 displayed potent and highly selective antiproliferative activity towards the MCF-7 human cancer cell line (IC50 = 5 nM). In the ER-binding assays, the lead conjugate 28 demonstrated potent ER competitive binding in ERα (IC50 value: 0.9 nM) and ERβ (IC50 value: 4.7 nM). Preliminary biochemical results also demonstrate that the lead conjugate 28 may exhibit pure antagonism. This series makes an important addition to the class of ER antagonists and may have potential applications in anticancer therapy.


Introduction
Estrogen receptors (ER), principally present as two main isoforms; ERα and ERβ, are found in abundance in female reproductive tissues such as the breast, uterus and ovary, while also found in bone, liver and brain tissue [1][2][3][4][5]. ERs can be overexpressed in tumour tissue and this provides a means to selectively target these tissues by both steroidal and non-steroidal ER ligands. ER ligands can be classified by their agonistic and antagonistic behaviour in the different ER-isoforms [3][4][5][6]. The term term selective estrogen receptor subtype modulator (SERSM) refers to the observation that a large number of reported ER-ligands have varying degrees of agonist/antagonistic behaviour towards the ERα and ERβ isoforms at the different tissue sites [3][4][5][6][7]. This leads to a complex action [1] where the benefits of a ligand at one ER-tissue site may be lessened by the negative effects the same ligand exerts at another ER-tissue site. For many decades, tamoxifen has been prescribed worldwide for the treatment of hormone-dependent breast cancer, (Figure 1). Tamoxifen displays antagonistic behaviour in breast tumour tissue; however this drug also displays agonistic behaviour on uterine tissue, which can lead to an increased risk of developing uterine cancer in postmenopausal women [3][4][5]. The other main concern in the use of tamoxifen is increased incidence of blood clots [8]. Breast cancer is often hormone dependent in its early stages of development. However as the disease progresses, the tumours can become less hormone dependent and difficult to treat [9,10]. For an effective treatment of hormone-dependent breast cancer, one goal would be to design an ER-ligand with no noticeable agonistic effects, thus displaying pure antagonistic properties. Various strategies attempt to improve the selectivity of chemotherapeutic agents by specifically targeting cancer cells and tumour environments [11,12]. Conjugates have been designed containing multiple pharmacophore elements or ligands, individually separated by a linker group, which aim to exert a synergistic and improved selective action on the target disease [13]. To date, a number of ER-targeting conjugates have been reported which attempt to exploit the high affinity and receptor selectivity of estrogen receptor ligands to deliver cytotoxic drugs to tumour cells [14][15][16][17][18]. In our investigation, antagonistic ER-ligands are key structural components utilised as the conjugate's targeting mechanism. In the present study, the ER-targeting antagonist endoxifen is linked via a covalent amide bond to a Combretastatin A-4 analogue-itself a possible antimitotic agent. We now investigate if the introduction of steric hindrance provided by the Combretastatin CA-4 amide fragment, would enhance the ER antagonistic effects of the endoxifen conjugate in the ER positive MCF-7 cells, possibly by interferance with Helix-12. It is hypothesised that the combination of an antagonistic ER-ligand and the Combretastatin CA-4 related acrylic acid antimitotic agent may produce a selective antiproliferative action on ER-dependent cancers.

Chemistry
All reagents were commercially available and were used without further purification unless otherwise indicated [19]. Tetrahydrofuran (THF) was distilled immediately prior to use from Na/Benzophenone under a slight positive pressure of nitrogen, toluene was dried by distillation from sodium and stored on activated molecular sieves (4 Å) and dichloromethane was dried by distillation from calcium hydride prior to use. Uncorrected melting points were measured on a Gallenkamp apparatus. Infra-red (IR) spectra were recorded as thin film on NaCl plates, or as Various strategies attempt to improve the selectivity of chemotherapeutic agents by specifically targeting cancer cells and tumour environments [11,12]. Conjugates have been designed containing multiple pharmacophore elements or ligands, individually separated by a linker group, which aim to exert a synergistic and improved selective action on the target disease [13]. To date, a number of ER-targeting conjugates have been reported which attempt to exploit the high affinity and receptor selectivity of estrogen receptor ligands to deliver cytotoxic drugs to tumour cells [14][15][16][17][18]. In our investigation, antagonistic ER-ligands are key structural components utilised as the conjugate's targeting mechanism. In the present study, the ER-targeting antagonist endoxifen is linked via a covalent amide bond to a Combretastatin A-4 analogue-itself a possible antimitotic agent. We now investigate if the introduction of steric hindrance provided by the Combretastatin CA-4 amide fragment, would enhance the ER antagonistic effects of the endoxifen conjugate in the ER positive MCF-7 cells, possibly by interferance with Helix-12. It is hypothesised that the combination of an antagonistic ER-ligand and the Combretastatin CA-4 related acrylic acid antimitotic agent may produce a selective antiproliferative action on ER-dependent cancers.

Chemistry
All reagents were commercially available and were used without further purification unless otherwise indicated [19]. Tetrahydrofuran (THF) was distilled immediately prior to use from Na/Benzophenone under a slight positive pressure of nitrogen, toluene was dried by distillation from sodium and stored on activated molecular sieves (4 Å) and dichloromethane was dried by distillation from calcium hydride prior to use. Uncorrected melting points were measured on a Gallenkamp
2.1.11. (E)-3-(4-Bromophenyl)-2-(3,4,5-trimethoxyphenyl)acrylic Acid 17 4-Bromobenzaldehyde (0.41 g, 2.21 mmol) and 3,4,5-trimethoxyphenylacetic acid (0.50 g, 2.21 mmol) were reacted following the general method above. Recrystallisation from ethanol yielded the acrylic acid 17 as fine yellow needles (0.36 g, 41%, m.p. 227-230˝C) [22]. 1    A mixture of the required acrylic acid (1 equivalent (eq.), 0.15 mmol), DCC (1 eq., 0.15 mmol, 0.03 g) and HOBt (1 eq., 0.15 mmol, 0.02 g) were suspended in 3 mL of anhydrous DCM and stirred for 10 min under a nitrogen atmosphere. The required silyl-protected endoxifen analogue, 9 (0.08 g, 0.15 mmol, 1 eq.) or 10 (0.10 g, 0.15 mmol, 1 eq.), was dissolved in 3 mL of anhydrous DCM and slowly added to the mixture via syringe. Reaction was allowed stir for 24-48 h. Reaction was monitored via TLC (DCM:MeOH, 4:1). The reaction mixture was diluted to 15 mL with anhydrous DCM and filtered to remove DCU. The filtrate was evaporated to dryness under reduced pressure. The residue was dissolved in 3 mL anhydrous THF and stirred under a nitrogen atmosphere. A solution of 0.1 M TBAF (2 equivalents) was added to the mixture and allowed stir for 24 h. The mixture was evaporated to dryness under reduced pressure. The residue was dissolved in DCM and washed with 10% HCl solution. The resulting organic phase was dried over sodium sulfate and evaporated to dryness under vacuum. The residue was purified via flash chromatography on silica gel (DCM:MeOH, 20:1) to yield a E/Z isomeric mixture of the products. The acrylic acid analogue 13 was reacted with the endoxifen derivative 9, following the general method above. The product 27 was afforded as a brown oil (103 mg, 94%), then changed to a semi-solid resin. 1   The acrylic acid analogue 13 was reacted with the endoxifen derivative 10, following the general method above,. The product 28 was afforded as a brown oil (104 mg, 92%), then changes to a semi-solid resin. 1   The acrylic acid analogue 14 was reacted with the endoxifen derivative 9, following the general method above. The product 29 was afforded as a brown oil (97 mg, 90%), then changes to a semi-solid resin. 1   The acrylic acid analogue 14 was reacted with the endoxifen derivative 10, following the general method above. The product 30 was afforded as a brown oil (104 mg, 94%), then changes to a semi-solid resin. 1   The acrylic acid analogue 23 was reacted with the endoxifen derivative 9, following the general method above. The product 31 was afforded as a brown oil (99 mg, 92%), then changes to a semi-solid resin. 1   The acrylic acid analogue 23 was reacted with the endoxifen derivative 10, following the general method above. The product 32 was afforded as a brown oil (99 mg, 90%), then changes to a semi-solid resin. 1 The acrylic acid analogue 25 was reacted with the endoxifen derivative 9, following the general method above. The product 33 was afforded as a brown oil (94 mg, 93%), then changes to a semi-solid resin. 1 The acrylic acid analogue 25 was reacted with the endoxifen derivative 10, following the general method above. The product 34 was afforded as a brown oil (95 mg, 92%), then changes to a semi-solid resin. 1

Lactate Dehydrogenase Assay for Measurement of Cytotoxicity
In this assay, the release of cytoplasmic lactate dehydrogenase (LDH) is used as a measure of cell lysis. MCF-7 and MDA-MB-231 cells were seeded at a density of 1ˆ10 4 cells/well in a 96-well plate and incubated for 24 h. The cells were then dosed with 2 µL volumes of the test compounds, over the concentration range 1 nM-50 µM. Cytotoxicity was determined using the CytoTox 96 nonradioactive cytotoxicity assay (Promega, Madison, WI, USA) following the manufacturer's protocol [32].

Estrogen Receptor Fluorescent Polarisation Assay
Competitive binding affinity experiments were carried out using purified baculovirus-expressed human ERα and ERβ and fluoromone, a fluorescein labelled estrogen ligand. Estrogen receptor binding ability of the selected compounds was investigated using ERα and ERβ fluorescence polarisation-based estrogen receptor competitive assay kits supplied by Invitrogen [33,34]. The assay was performed usibg a protocol described by the manufacturer. The assay allows for high throughput screening of potential ER-subtype ligands. The selected compounds were screened using both the ERα and ERβ competitive assay kits. The protocol for carrying out the assay is similar for both ER subtypes. Principally, the main difference between the kits relates to the functional receptor concentration and the specific activity of the different ERs [33,34].

Ishikawa Cell Line Study
The Ishikawa assay is used to measure estrogen stimulation of alkaline phosphatase enzyme activity (AlkP) by the Ishikawa cell line of human endometrial adenocarcinoma cells. The Ishikawa assay provides a measure of the agonist activity of a compound. The assay was carried out on the lead conjugate, 28. The assay was carried out following the method of Littlefield et al. [35]. The batch of Ishikawa cells were obtained as a gift from Professor R. Hochberg-who developed the alkaline phosphatase assay in Yale University, CT, USA. Tamoxifen was used as a reference compound in the assay.

NCI One-Dose and Five-Dose Screen Output
The one-dose screen output is reported as a mean graph of the percent growth of treated cells and is similar in appearance to mean graphs generated in the 5-dose assay. The value reported for the one-dose assay is growth relative to the no-drug control and relative to the time zero number of cells. The one-dose assay allows detection of growth inhibition (values between 0 and 100) and lethality (values less than 0). For example, a value of 100 means no growth inhibition. A value of 40 would mean 60% growth inhibition. A value of zero means no net growth over the course of the experiment. A value of´40 would mean 40% lethality. A value of´100 means all cells are dead. The results from the one-dose screen for 28 were manually entered into the COMPARE analysis software via an on-line submission form [36]. The results from the COMPARE analysis are retrievable on-line by searching using the relevant JobID reference number. The COMPARE analysis was run on a database of common anti-cancer agents (JobID: 37472) and the larger more comprehensive database including natural products and other submitted agents (JobID: 37473). Similarly, the results from the five-dose screen for 28 were also manually entered into the COMPARE analysis software via an on-line submission form. The COMPARE analysis was run on a database of common anti-cancer agents (JobID: 37885) and the larger more comprehensive database including natural products and other submitted agents (JobID: 37886).

Molecular Modelling
The lowest energy conformer produced [37] using MACROMODEL [38] was used to generate an ensemble of low energy conformations of 28 in OMEGA [39]. Fifty conformers were generated for the lead conjugate 28 using default parameters and saved as a .pdb file. The resulting .pdb file generated by OMEGA was then utilised by FRED [40] to dock and score the different compound conformers. The protein used to dock the conformers was 3ERT [41] (containing 4-hydroxytamoxifen bound in the human ERα LBD) and 1QKN [42] (containing raloxifene bound in rat ERβ LBD). The active site was defined by a three-dimensional box incorporating the 4-hydroxytamoxifen or raloxifene bound ligand. This box was also extended by five angstroms in each dimension to create additional space to allow for the docking procedure. Each conformer was docked and scored using three functions: Piecewise Linear Potential (PLP), Chemgauss and an updated version, Chemgauss2 [40].
PLP is a heavy atom scoring function; all potentials are based on distances from heavy atom centers (i.e., hydrogen position is irrelevant, although the presence or absence of hydrogen is not, as it can affect the atom typing). PLP recognises atom types such as hydrogen bond donors (i.e., primary and secondary amines), hydrogen bond acceptors (i.e., oxygen and nitrogen atoms with no bound hydrogens), hydroxyl groups (treated as both acceptors and donors) and non-polar entities (i.e., carbon, halogens and nitrogen or sulfur with more than two attached hydrogens). The Chemgauss scoring function combines the Shapegauss scoring function with additional potentials near specific functional groups. The Shapegauss scoring function represents all atoms as smooth Gaussian functions. A pairwise potential between ligand and protein atoms is applied that attempts to maximize their surface contact and minimize their volume overlap. Therefore, the potential is most favourable when the atoms are touching but not overlapping. A correction term is then applied to further penalize atoms that significantly overlap the protein. A consensus of the separate scoring functions is determined and the conformers are ranked accordingly.
The crystal structure of raloxifene, an antagonistic ligand, in rat ERβ isoform (pdb: 1QKN) was used due in this study due to the lack of reported determined co-crystallised antagonist ligands in the human ERβ isoform.

Synthesis of Endoxifen-Combretastatin Conjugates
Many ER-ligand conjugates reported in the literature contain an agonistic ER-ligand analogue such as estrogen, in their conjugate structure [14,15]. As the goal in our investigation is to develop ER-antagonistic conjugates, possessing minimal agonist activity, only antagonistic ER-ligands were incorporated in the conjugate structural backbone. Endoxifen 11 which together with 4-hydroxytamoxifen is a significant metabolite of tamoxifen, was chosen as a suitable ER-ligand candidate for this study based on its potent ER-binding affinity and antiestrogenic properties [43]; it was also effective in degrading the estrogen receptor [44] (Figure 1). It has also been shown to inhibit aromatase [45] in the MCF-7 human cancer cell line [46]. The structurally related hydroxyendoxifen analogue 12, previously detected as a metabolite of tamoxifen [47], was also investigated as a potential ER ligand for conjugate design (Figure 1).
A modified multi-step route to the protected endoxifen scaffolds 9 and 10 was developed and is shown in Scheme 1. Initially, the phenolic 4,4 1 -dihydroxybenzophenone (1) was monoprotected as the tert-butyldimethylsilyl-ether (2); the 4-hydroxypropiophenone (3) was similarly protected to afford (4b). The benzophenone (2) and propiophenone starting materials (4a) and (4b) were coupled via a zinc/titanium tetrachloride/tetrahydrofuran McMurray reaction system to afford alkene products 5 and 6 in high yields (93%-98%) containing the triarylethylene ring motif predominant OH directed E-isomer [48]. A bromoethylation reaction was carried out using an excess of dibromoethane in the presence of a phase-transfer catalyst((nBu) 4 NHSO 4 ) in basic conditions in order to introduce a bromoethylether functionality at the hydroxyl group position on the triarylethylene backbone in moderate yields (52%-54%). Methylamine undergoes reaction with the relevant bromide analogues 7 and 8 in a sealed tube to form the silyl-protected endoxifen 9 [16] and the disilylated endoxifen analogue 10 in moderate to high yields (55%-93%). The silyl ether protecting groups were removed using TBAF to afford the endoxifen 11 and hydroxyendoxifen 12 in high yields (80%-93%). Most of the E/Z isomeric mixture ratios of the analogues synthesised were calculated based on the basic side chain OCH 2 or NCH 2 signals. Unambiguous assignment of E/Z stereochemistry is confirmed using through Nuclear Overhauser Effect (NOE) NMR. In the 1 H-NMR spectrum of the silyl ether 9 (Z:E ratio 1:1.3) the E-isomer CH 2 N signals are observed as a broad triplet at 2.99 ppm while the Z-isomer CH 2 N signals are observed as a triplet further upfield at 2.89 ppm. The E-isomer CH 2 O signals are observed as a triplet at 4.11 ppm(J = 5 Hz) while the Z-isomer CH 2 O signals are observed as a triplet further upfield at 3.95 ppm(J = 5 Hz). The relative chemical shifts assigned for the OCH 2 and NCH 2 signals for the protons in the basic side chain in the isomeric mixtures are in agreement with reported values for similar compounds [29]. The spectral data confirm that the trans-isomer OCH 2 (and NCH 2 ) signals are found further downfield when compared with the cis-isomer. Previous studies have demonstrated that 4-hydroxy substituted triarylethylenes such as endoxifen may isomerise under physiological conditions and have little impact on ER binding activity [49,50]. Combretastatin A-4 is an important lead compound in drug development due to its potent antimitotic activity and ability to inhibit the depolymerisation of tubulin (Figure 1) [51]. To date, much work has been carried out in developing combretastatin analogues with potential anticancer applications [52]. The conjugation of combretastatin A4 analogues on steroidal scaffolds and their proapoptotic effects in MCF-7 cells has been recently reported [53]. In this study, a selection of combretastatin acrylic acid derivatives were synthesised using a reported Perkin condensation reaction route [21]. The common structural motif amongst the acrylic acid derivatives synthesised was the presence of a carboxylic acid group at either carbon position of the double bond between the two ring systems of the combretastatin core structure. Importantly, the carboxylic acid group allows for further chemical manipulation, such as the formation of amide and ester linkages, which is of interest in our conjugation strategy.
The combretastatin acrylic acids chosen for synthesis were selected based on the biochemical activity available for the "parent" combretastatin analogues [52]-and which differed in structure only by the absence of the acrylic acid's carboxylic acid functional group [22,[54][55][56][57][58][59][60][61][62][63][64][65]. By varying the benzaldehyde and phenylacetic acid in the Perkin condensation reaction, a series of combretastatin acrylic acid analogues 13-24 were prepared in yields of 36%-71% (see Scheme 2). In our investigation, combretastatin A-4 was used as a comparison standard for our biochemical evaluation and was synthesised according to a Wittig reaction route reported by Pettit et al. [20]. Combretastatin A-4 is an important lead compound in drug development due to its potent antimitotic activity and ability to inhibit the depolymerisation of tubulin (Figure 1) [51]. To date, much work has been carried out in developing combretastatin analogues with potential anticancer applications [52]. The conjugation of combretastatin A4 analogues on steroidal scaffolds and their proapoptotic effects in MCF-7 cells has been recently reported [53]. In this study, a selection of combretastatin acrylic acid derivatives were synthesised using a reported Perkin condensation reaction route [21]. The common structural motif amongst the acrylic acid derivatives synthesised was the presence of a carboxylic acid group at either carbon position of the double bond between the two ring systems of the combretastatin core structure. Importantly, the carboxylic acid group allows for further chemical manipulation, such as the formation of amide and ester linkages, which is of interest in our conjugation strategy.
The combretastatin acrylic acids chosen for synthesis were selected based on the biochemical activity available for the "parent" combretastatin analogues [52]-and which differed in structure only by the absence of the acrylic acid's carboxylic acid functional group [22,[54][55][56][57][58][59][60][61][62][63][64][65]. By varying the benzaldehyde and phenylacetic acid in the Perkin condensation reaction, a series of combretastatin acrylic acid analogues 13-24 were prepared in yields of 36%-71% (see Scheme 2). In our investigation, combretastatin A-4 was used as a comparison standard for our biochemical evaluation and was synthesised according to a Wittig reaction route reported by Pettit et al. [20]. From the panel of combretastatin type acrylic acids 13-24 initially prepared, a subset of the more potent compounds was selected for the synthesis of the direct amide conjugates with endoxifen 11 and hydroxyendoxifen 12. The amine functional group of the silyl-protected ER ligands 9 and 10 and the carboxylic acid functional group of the combretastatin acrylic acid analogues 13-24 can undergo coupling reactions using DCC and HOBt, forming an amide linkage to afford silyl-ether protected conjugated compounds in high yields (See Scheme 3). The silyl-ether protecting groups were then removed using TBAF to afford the phenolic conjugates 27-46 as~1:1 (E/Z) isomeric mixtures in high yield (88%-94%) following chromatographic purification. In the 1 H NMR spectrum the characteristic protons of the ethyl group are observed in the region 0.90-0.96 ppm (CH 3 ) and 2.40-2.50 ppm (CH 2 ), the amine methyl group is found between 3.04 and 3.25 ppm, while the methylene protons of the basic side chain are identified at 3.53-4.20 ppm. The amide compounds 47-53 were also prepared by reaction of the acrylic acids 13, 14, 21 and 23 with pyrrolidine and piperidine using the Mukaiyama reagent (2-chloro-1-methylpyridinium iodide) as the coupling reagent. The amines 54 and 55 were obtained by reduction of the corresponding nitro compounds 50 and 51 using iron in hydrochloric acid.  Table 2); (iii) Pyrrolidine or piperidine, 2-chloro-1-methylpyridinium iodide, CH2Cl2, Et3N, 20 °C, 1 h; (iv) Fe, HCl, CH3CO2H, EtOH, reflux, 12 h.

Antiproliferation and Cytotoxicity Studies
The ability of the compounds 27-46 synthesised to inhibit the proliferation of the human breast tumour MCF-7 cell line was investigated using a standard MTT cell viability assay while the compounds were concurrently tested to assess the extent of their cytotoxicity using a LDH assay [66]. The MCF-7 is an ER-positive human cancer cell line; where ER is overexpressed in the cell line [67]. Selected conjugates were also evaluated using ER-negative MDA human cancer cell-line in order to assess any possible ER-selectivity of the conjugates. The IC50 value obtained for the control 26 (CA4) in this assay is 0.008 µM for MCF-7 and is in good agreement with the reported values for CA4 using the MTT assay on human MCF-7 breast cancer cell lines [59,68,69]; the IC50 value obtained   Table 2); (iii) Pyrrolidine or piperidine, 2-chloro-1-methylpyridinium iodide, CH 2 Cl 2 , Et 3 N, 20˝C, 1 h; (iv) Fe, HCl, CH 3 CO 2 H, EtOH, reflux, 12 h.

Antiproliferation and Cytotoxicity Studies
The ability of the compounds 27-46 synthesised to inhibit the proliferation of the human breast tumour MCF-7 cell line was investigated using a standard MTT cell viability assay while the compounds were concurrently tested to assess the extent of their cytotoxicity using a LDH assay [66]. The MCF-7 is an ER-positive human cancer cell line; where ER is overexpressed in the cell line [67]. Selected conjugates were also evaluated using ER-negative MDA human cancer cell-line in order to assess any possible ER-selectivity of the conjugates. The IC 50 value obtained for the control 26 (CA4) in this assay is 0.008 µM for MCF-7 and is in good agreement with the reported values for CA4 using the MTT assay on human MCF-7 breast cancer cell lines [59,68,69]; the IC 50 value obtained for the control endoxifen 11 in this assay is 0.029 nM while the reported IC 50 value for endoxifen in the MCF-7 cell line is 50 nM [70].
In general, combretastatin acrylic acid derivatives have a lower antiproliferative potency compared to the lead compound, combretastatin A-4 (CA4), 26 (IC 50 = 8 nM), (see Table 1). The acrylic acid 13 displayed the highest antiproliferative action in the series, with IC 50 value of 0.120 µM while all the analogues tested showed negligible cytotoxic affects in the LDH assay, (0% cell death at 10 µM). The substitution pattern in the A and B rings of 13 are similar to that of the potent combretastatin CA4. It is interesting that the compound 23, in which the A ring having the 3,4,5-trimethoxyphenyl substituent is positioned on the carbon β to the acrylic acid is inactive with IC 50 > 50 µM. The acrylic acids 14, 16, 21 and the ester 24 all demonstrated IC 50 values less than 10 µM against MCF-7 cell line. The antiproliferative activity of acrylic acids has been previously reported [23]. The acrylic acid secondary amides 47-55 prepared from pyrrolidine and piperidine were also evaluated for antiproliferative activity in the MCF-7 cell line ( Table 1). The compounds were found to be of low potency, e.g., the amide 49 (IC 50 = >50 µM in comparison with the acrylic acid 13 (IC 50 = 0.120 µM) which suggested that a more bulky amide such as endoxifen may be required for activity in MCF-7 cells. for the control endoxifen 11 in this assay is 0.029 nM while the reported IC50 value for endoxifen in the MCF-7 cell line is 50 nM [70]. In general, combretastatin acrylic acid derivatives have a lower antiproliferative potency compared to the lead compound, combretastatin A-4 (CA4), 26 (IC50 = 8 nM), (see Table 1). The acrylic acid 13 displayed the highest antiproliferative action in the series, with IC50 value of 0.120 µM while all the analogues tested showed negligible cytotoxic affects in the LDH assay, (0% cell death at 10 µM). The substitution pattern in the A and B rings of 13 are similar to that of the potent combretastatin CA4. It is interesting that the compound 23, in which the A ring having the 3,4,5-trimethoxyphenyl substituent is positioned on the carbon β to the acrylic acid is inactive with IC50 > 50 µM. The acrylic acids 14, 16, 21 and the ester 24 all demonstrated IC50 values less than 10 µM against MCF-7 cell line. The antiproliferative activity of acrylic acids has been previously reported [23]. The acrylic acid secondary amides 47-55 prepared from pyrrolidine and piperidine were also evaluated for antiproliferative activity in the MCF-7 cell line ( Table 1). The compounds were found to be of low potency, e.g., the amide 49 (IC50 = >50 µM in comparison with the acrylic acid 13 (IC50 = 0.120 µM) which suggested that a more bulky amide such as endoxifen may be required for activity in MCF-7 cells.   Table 2 and thus definitively determine whether the antiproliferative effects observed had a ER-mediated component or whether these effects were solely mediated through the combretastatin component of the conjugate.

Estrogen Receptor Binding
The receptor binding affinity for the lead conjugate 28 was investigated in both ERα and ERβ using a fluorescence polarisation based competitive binding assay. Competitive binding affinity experiments were carried out using purified baculovirus-expressed human ERα and ERβ and fluoromone, a fluorescein labelled estrogen ligand [33,34]. The endogenous ligand β-estradiol was used as the positive control in the experiments. The lead conjugate displayed potent binding in both ER isoforms (IC 50 value for ERα: 0.9 nM; IC 50 value for ERβ: 4.7 nM). These binding values are greater than the endogenous ligand estradiol, 4-hydroxytamoxifen, endoxifen 11 and significantly greater than that of the parent ER-ligand, hydroxyendoxifen 12 (ERα: IC 50 44.1 nM; ERβ: IC 50 39.7 nM) (see Table 4). Table 4. Estrogen receptor binding affinities for compounds 11, 12 and 28 a . a Mean IC50 values of compounds for their antiproliferative effects and percent cytotoxicity on a human MDA breast cancer cell line. b IC50 values are half maximal inhibitory concentrations required to block the growth stimulation of MDA-MB-231 cells. Values are an average of at least three experiments performed in triplicate with typical standard errors below 15%. c Lactate Dehydrogenase assay: Following treatment of the cells, the amount of LDH was determined using LDH assay kit from Promega. Data is presented as % cell lysis at compound concentration of 10 µM [66].

Estrogen Receptor Binding
The receptor binding affinity for the lead conjugate 28 was investigated in both ERα and ERβ using a fluorescence polarisation based competitive binding assay. Competitive binding affinity experiments were carried out using purified baculovirus-expressed human ERα and ERβ and fluoromone, a fluorescein labelled estrogen ligand [33,34]. The endogenous ligand β-estradiol was used as the positive control in the experiments. The lead conjugate displayed potent binding in both ER isoforms (IC50 value for ERα: 0.9 nM; IC50 value for ERβ: 4.7 nM). These binding values are greater than the endogenous ligand estradiol, 4-hydroxytamoxifen, endoxifen 11 and significantly greater than that of the parent ER-ligand, hydroxyendoxifen 12 (ERα: IC50 44.1 nM; ERβ: IC50 39.7 nM) (see Table 4).

NCI 60 Cell Line Screen
On the basis of potency, compound 28 was evaluated in the National Cancer Institute (NCI, Bethesda, MD, USA) Division of Cancer Treatment and Diagnosis(DCTD)/Developmental Therapeutics Programme(DTP) in which the activity of the compound is determined using a 60-cell line screen facility of different cancer cell lines of diverse tumour origin [71]. Compound was tested for inhibition of growth(GI 50 ) and cytotoxicity(LC 50 ). These studies were performed at the NCI as part of their drug screening programme. Initially, the compound was evaluated against the 60 cell lines at a single dose of 10 µM; if significant growth inhibition was exhibited the compound was evaluated against the 60 cell panel at a further five concentration levels, 0.01-100 µM, (see Table 5). In the one-dose screen, 28 displayed low lethality and a mean growth inhibition value of 98.73% over the 60 cell lines. 28 displayed very high growth inhibition in the cell lines of non-small cell lung cancer NCI-H23 (97%) and NCI-H460 (98%); colon cancer HCT-116 (98%); breast cancer MCF-7 (94%) and MDA-MB-435 (91%); ovarian cancer OVCAR-8 (97%) and SK-OV-3 (95%); leukaemia RPMI-8226 (91%); renal cancers ACHN (96%), CAKI-1 (91%), RXF 393 (95%) and SN12C (90%); melanoma SK-MEL-2 (97%) and CNS cancer U251 (93%). The compound caused between 80%-89% growth inhibition in a further 14 cell lines and was very toxic to the non-small cell lung cancer NCI-H226 (73% lethality) and melanoma SK-MEL-5 (60% lethality) cell lines. In the five-dose screen, 28 displayed low micromolar GI 50 (IC 50 ) values for the majority of the 60 cancer cell lines. However, 28 demonstrated a high selectivity towards MCF-7 breast cancer with a GI 50 (IC 50 ) value of 9.5 nM and a LC 50 value greater than 50 µM for this cell line. The mean GI 50 value compound 28 across all 60 panel cell lines is 1.45 µM, (log GI 50 =´5.79). As 28 has been shown to be a high-affinity ER-binding ligand, the high specificity towards MCF-7 cells is most probably due to its selective antagonistic effects on the over-expressed ER within this cell-line. This is an impressive and promising result as it confirms the suitability of the selected prototypes and project strategy for the possible therapeutic application of the ER-conjugates synthesised in the study. The LC 50 Table 5).
The NCI provide the pattern recognition algorithm COMPARE [72,73]. The unique complexity of a 60-cell line dose response produced by a given compound results in a biological response pattern which can be utilized in pattern recognition algorithms. Using the COMPARE algorithms, it is possible to assign a putative mechanism of action for the screened compound, or to determine that the response pattern is unique and not similar to that of any of the standard prototype compounds included in the NCI database. In addition, following characterization of various cellular molecular targets in the 60-cell lines, it may be possible to select compounds most likely to interact with a specific molecular target. Generally, a correlation coefficient greater than 0.6 is considered a positive correlation [74]. For the five-dose screen COMPARE analyses, the highest correlation coefficients achieved for compound 28, from both the common anticancer agent database and the comprehensive database, were 0.934 in relation to the potent antiviral and anticancer agent, Didemnin B [75] which selectively induces apoptosis through dual inhibition of PPT1 and EEF1A1, 0.879 for the antileukemic agent and protein synthesis inhibitor, Bruceantin [76] and 0.430 in relation to the antimitotic/anti-tumour agent Rhizoxin 47 [77,78]. Correlation values are Pearson correlation coefficients based on GI 50 mean graphs. These correlation values demonstrate a high similarity of activity and suggest a common mechanism of action between the agents. The antiproliferative activity observed for the conjugate compound 28 indicated that there is a significant therapeutic window between the concentration required for cancer cell growth inhibition and the concentration that is toxic to MCF-7 cells.

Ishikawa Cell Line Study
The Ishikawa assay is used to measure estrogen stimulation of alkaline phosphatase enzyme activity (AlkP) by the Ishikawa cell line of human endometrial adenocarcinoma cells [35]. The Ishikawa assay provides a measure of the agonist activity of a compound and was carried out on the lead conjugate, 28. Tamoxifen was used as a reference compound in the assay. The effect of the compounds on estrogen stimulation within the Ishikawa cells can be seen in Figure 2, which is representative of an experiment that was carried out three times. Tamoxifen displays some estrogenic activity at higher concentrations. When Tamoxifen is dosed with a 1 nM estradiol spike, there is estrogen stimulation at low concentration while a reduction occurs as the concentration of Tamoxifen is increased. The conjugate was dosed in duplicate. The conjugate 28 is similar to Tamoxifen when dosed individually. However, the conjugate appears not to display any estrogenic activity at higher concentrations, which is desirable. This may suggest that the conjugate is acting as a full antagonist. When 28 was dosed as a mixture with estradiol, there is a reduction in estrogenic stimulation at low concentrations and a more pronounced reduction in estrogen stimulation over the range of concentrations when compared with the Tamoxifen mixture. This preliminary result is promising as it demonstrates that the conjugate 28 does not display estrogenic stimulation in the cell line and shows reduced estrogenic stimulation compared to Tamoxifen over the concentration range investigated.
Biomedicines 2016, 4, 15 28 of 34 activity at higher concentrations, which is desirable. This may suggest that the conjugate is acting as a full antagonist. When 28 was dosed as a mixture with estradiol, there is a reduction in estrogenic stimulation at low concentrations and a more pronounced reduction in estrogen stimulation over the range of concentrations when compared with the Tamoxifen mixture. This preliminary result is promising as it demonstrates that the conjugate 28 does not display estrogenic stimulation in the cell line and shows reduced estrogenic stimulation compared to Tamoxifen over the concentration range investigated.

Stability Studies
The stability of the selected target compounds 27, 31 and 32 was evaluated in phosphate buffer at pH values in the range 4-9 and the half-life was determined to be greater than 24 h for each compound at these pH values. The compounds were found to stable without any significant degradation of the conjugates suggesting that the combretastatin fragment remains intact at

Stability Studies
The stability of the selected target compounds 27, 31 and 32 was evaluated in phosphate buffer at pH values in the range 4-9 and the half-life was determined to be greater than 24 h for each compound at these pH values. The compounds were found to stable without any significant degradation of the conjugates suggesting that the combretastatin fragment remains intact at physiological pH; and indicating that the combretastatin component of the conjugate has a role in displacing helix-12 of the ER resulting in the potent (and possibly pure) antagonistic activity.

Molecular Modelling
The intricate molecular basis of estrogen receptor agonism and antagonism has been well studied [1,[79][80][81]. Phenol groups, a common structural motif to many ER ligands, take part in direct hydrogen bonding with the same key residues in the ER (i.e., Glu353, Arg 394 and His524 in human ERα). Within the ER ligand-binding domain (LBD), ligand recognition is achieved through a combination of specific hydrogen bonds and the complementarity values of the binding cavity with the relevant ligands non-polar character. Antagonistic ligands such as Tamoxifen and 4-hydroxytamoxifen contain a basic side chain group too large to be accommodated in the binding cavity, resulting in the displacement of the Helix-12 in the ER protein structure. In particular, the positioning of Helix-12 in the ER is key to the recruitment of coactivators/corepressors and determining the overall nature of the ligand effect [81].
The interactions of compound 28 was modelled in both ERα and ERβ isoforms in order to rationalise the biochemical data obtained for this compound. The modelling strategy protocol involved calculating the appropriate protonation state at physiological pH using MARVIN [82]; the determination of the lowest energy conformer using MACROMODEL [38]; the generation of an ensemble of low-energy conformers using OMEGA [39], followed by the docking and scoring of these conformers using FRED [40]. Graphical manipulations were carried out using DS VISUALIZER [83]. High-ranking docking solutions were investigated-noting favourable interactions between the conjugate and the residues in the ER isoform ligand binding domains.
The top-ranking docking solution in human ERα for compound 28 is shown in Figure 3. For comparison, the 4-hydroxytamoxifen bound ligand (reference pdb: 3ERT) [41] is displayed in yellow while the docked confirmations are coloured by element type (carbon = grey, hydrogen = white, oxygen = red, nitrogen = blue). In Figure 3, the Glu353 and Arg394 residues form hydrogen bond interactions with the phenol group, similar to that of 4-hydroxytamoxifen. In this docked solution the additional phenol group of 28 is found 3.9 Å away from the His524 residue and does not form a strong H-bond. The Asp351 residue is located 3.7 Å away from the amide nitrogen of 28, which would translate to a weak interaction. The Cys530 residue forms a H-bond with the C-4 methoxy group on the A-ring of the combretastatin fragment as illustrated in Figure 3. The increased bulk of the ligand side chain of the conjugate 28 may displace Helix-12 to a significant extent and could explain the antagonistic activity and impressive binding affinity of this lead compound 28 (0.9 nM in ERα; 4.7 nM in ERβ).
The top-ranking docking solutions in rat ERβ for compound 28 is shown in Figure 4. For comparison, the raloxifene bound ligand (reference pdb: 1QKN) [42] is displayed in yellow while the docked confirmations are coloured by element type (carbon = grey, hydrogen = white, oxygen = red, nitrogen = blue). In Figure 4, the residues Glu260 and Arg301 form favourable hydrogen bond interactions with the phenol group, similar to that of 4-hydroxytamoxifen. The additional group present on 28 forms an additional hydrogen bond interaction with the His430 residue. There is good agreement/overlap of the raloxifene core and the endoxifen moiety of 28. A methoxy group present on the combretastatin moiety of 28 forms another favourable hydrogen bond through a water molecule with the Asp258 residue within the LBD. This interaction could help "lock in" the ligand within the cavity site, resulting in the good binding affinity. Additional waters unaccounted for in this determined crystal structure may introduce further favourable hydrogen bond interactions with the other methoxy groups of the combretastatin moiety through a network of water molecules. The alkyl groups of Ile328 are in close proximity to a phenyl group present on the endoxifen moiety, which could produce a lipophilic-lipophilic interaction. The bulky side chain of 28 would have a greater degree of rigidity due to the presence of the amide linkage and may displace Helix-12 significantly in ERβ to produce antiproliferative effects observed. interactions with the phenol group, similar to that of 4-hydroxytamoxifen. In this docked solution the additional phenol group of 28 is found 3.9 Å away from the His524 residue and does not form a strong H-bond. The Asp351 residue is located 3.7 Å away from the amide nitrogen of 28, which would translate to a weak interaction. The Cys530 residue forms a H-bond with the C-4 methoxy group on the A-ring of the combretastatin fragment as illustrated in Figure 3. The increased bulk of the ligand side chain of the conjugate 28 may displace Helix-12 to a significant extent and could explain the antagonistic activity and impressive binding affinity of this lead compound 28 (0.9 nM in ERα; 4.7 nM in ERβ). The top-ranking docking solutions in rat ERβ for compound 28 is shown in Figure 4. For comparison, the raloxifene bound ligand (reference pdb: 1QKN) [42] is displayed in yellow while the docked confirmations are coloured by element type (carbon = grey, hydrogen = white, oxygen = red, nitrogen = blue). In Figure 4, the residues Glu260 and Arg301 form favourable hydrogen bond interactions with the phenol group, similar to that of 4-hydroxytamoxifen. The additional group present on 28 forms an additional hydrogen bond interaction with the His430 residue. There is good agreement/overlap of the raloxifene core and the endoxifen moiety of 28. A methoxy group present on the combretastatin moiety of 28 forms another favourable hydrogen bond through a water molecule with the Asp258 residue within the LBD. This interaction could help "lock in" the ligand within the cavity site, resulting in the good binding affinity. Additional waters unaccounted for in this determined crystal structure may introduce further favourable hydrogen bond interactions with the other methoxy groups of the combretastatin moiety through a network of water molecules. The alkyl groups of Ile328 are in close proximity to a phenyl group present on the endoxifen moiety, which could produce a lipophilic-lipophilic interaction. The bulky side chain of 28 would have a greater degree of rigidity due to the presence of the amide linkage and may displace Helix-12 significantly in ERβ to produce antiproliferative effects observed.  Highest-ranking docking orientation and predicted receptor interactions for 28 (ball-and-stick representation) in ERβ compared to the crystal solution for 4-hydroxytamoxifen (yellow) (pdb 1QKN) [42]. (H-bond interactions are illustrated as broken lines).

Conclusions
A series of prototype ER-ligand conjugates 27-46 were successfully synthesised incorporating an endoxifen-combretastatin hybrid scaffold with potential SERM properties. A number of these novel compounds displayed potent antiproliferation activity in the MCF-t human breast cancer cell line with low cytotoxicity values. The conjugate 28 in the series was the most promising compound Highest-ranking docking orientation and predicted receptor interactions for 28 (ball-and-stick representation) in ERβ compared to the crystal solution for 4-hydroxytamoxifen (yellow) (pdb 1QKN) [42]. (H-bond interactions are illustrated as broken lines).

Conclusions
A series of prototype ER-ligand conjugates 27-46 were successfully synthesised incorporating an endoxifen-combretastatin hybrid scaffold with potential SERM properties. A number of these novel compounds displayed potent antiproliferation activity in the MCF-t human breast cancer cell line with low cytotoxicity values. The conjugate 28 in the series was the most promising compound in the study, with potent antiproliferative activity in MCF-7 human breast cancer cell line (IC 50 : 5 nM), low cytotoxicity and impressive ER competitive binding IC 50 values (ERα: IC 50 0.9 nM; ERβ: IC 50 4.7 nM). A number of these conjugate compounds also display activity in the ER negative cell line MDA-MB-321 at low micromolar and sub-micromolar concentrations, e.g., compound 34 (IC 50 = 0.68 µM). Preliminary stability studies show that these conjugates do not degrade significantly at physiological pH values. Interestingly, from a structural biology standpoint, this result indicates that the ER can tolerate a large substituent such as the combretastatin analogue, at the basic side chain of the triarylethylene ER ligand scaffold without detrimental effect on the overall antiproliferative and ER binding characteristics and biochemical activity. The lead conjugate compounds 27 and 28 are currently under further investigation by our research group to further explore its mechanism of action and potential applications in the development of useful ER ligands. These compounds demonstrate a novel and interesting class of ER ligand and may have potential future application as medicinal agents for anticancer therapy.