Atypical McMurry Cross-Coupling Reactions Leading to a New Series of Potent Antiproliferative Compounds Bearing the Key [Ferrocenyl-Ene-Phenol] Motif

In the course of the preparation of a series of ferrocenyl derivatives of diethylstilbestrol (DES), in which one of the 4-hydroxyphenyl moieties was replaced by a ferrocenyl group, the McMurry reaction of chloropropionylferrocene with a number of mono-aryl ketones unexpectedly yielded the hydroxylated ferrocenyl DES derivatives, 5a–c, in poor yields (10%–16%). These compounds showed high activity on the hormone-independent breast cancer cell line MDA-MB-231 with IC50 values ranging from 0.14 to 0.36 µM. Surprisingly, non-hydroxylated ferrocenyl DES, 4, showed only an IC50 value of 1.14 µM, illustrating the importance of the hydroxyethyl function in this promising new series. For comparison, McMurry reactions of the shorter chain analogue chloroacetylferrocene were carried out to see the difference in behaviour with mono-aryl ketones versus a diaryl ketone. The effect of changing the length of the alkyl chain adjacent to the phenolic substituent of the hydroxylated ferrocenyl DES was studied, a mechanistic rationale to account for the unexpected products is proposed, and the antiproliferative activities of all of these compounds on MDA-MB-231 cells lines were measured and compared. X-ray crystal structures of cross-coupled products and of pinacol-pinacolone rearrangements are reported.

hexestrol are also known [25]. In this new series, our goal was to compare the effect of chloroethyl (as in compound 3b) and ethyl (as in compound 4) groups on the antiproliferative activity of these compounds, as had been done previously for molecules 1 and 2 [26].
Somewhat surprisingly however, attempts to prepare the aforementioned chlorinated compounds, via McMurry coupling of chloropropionylferrocene and 4-hydroxyphenyl-alkyl ketones of varying alkyl chain length, led instead to hydroxyethyl derivatives (5a-c) resulting from the replacement of chlorine atom by a hydroxyl group; in addition, compounds resulting from pinacol-to-pinacolone rearrangement were also isolated. We here report the results of these atypical McMurry coupling reactions, and make comparison with the more usual reactions seen when using 4,4'dihydroxybenzophenone as the purely organic ketone counterpart.

Synthetic Aspects
Following the synthetic procedure successfully used for 2, the chloroethyl derivative of ferrocidiphenol, 1 [26], we first performed a McMurry coupling between chloropropionylferrocene, 6 [27], and 4-hydroxypropiophenone, 7b (Scheme 2). Surprisingly, only the hydroxyl compound 5b was isolated, albeit in low yield (16%). The crude oil showed a mixture of inseparable compounds, 5b being the only compound isolated by chromatographic techniques in pure form. This was attributable to its higher polarity induced by the hydroxyl group, as compared to the other by-products. Moreover, the expected chlorinated compound 3b could not be detected by mass spectrometry in the reaction mixture. Again, when 4-hydroxypropiophenone, 7b, was replaced by 4-hydroxyacetophenone, 7a, or 4-hydroxybutyrophenone, 7c, in the reaction with chloropropionylferrocene, 6, only the hydroxyl compounds 5a and 5c were isolated, in yields of 10% and 15%, respectively. This contrasts with the previously observed reaction of 4,4'-dihydroxybenzophenone, 7e, with 6 that furnished 2 (in 47% yield) [26] with no formation of hydroxylated compound 5e.

Scheme 2.
McMurry coupling reactions of chloropropionylferrocene, 6. Two isomers, Z and E, were formed, and the designation E refers to the isomer whereby the ferrocenyl group is trans relative to the phenol.  These observations are apparently specific to the McMurry reaction between 6 and mono-aryl ketones, since the exchange of chlorine to hydroxyl was found only in these cases. The difference is almost certainly not a steric effect, although electronic effects such as π-stacking between a hydroxyphenyl group and the ferrocenyl moiety during the formation of the intermediate pinacolate [28] may play a role. The lower yields of cross-coupled product when using mono-aryl ketones 7a-c with 6, rather than the diaryl ketone 7e, may be a result of increased homocoupling and also of pinacol-to-pinacolone rearrangement of intermediate titanium pinacolates. It is known that in McMurry reactions with two mono-aryl ketones there is less selectivity for cross-coupling than when using species as different as a mono-aryl and a diaryl-ketone. In our previous studies involving McMurry reactions between a mono-aryl and a diaryl ketone, the cross-coupling yields were generally quite high, homo-coupling was low, and pinacol rearrangement products were barely detectable. However, in the reactions of diaryl ketones and [3] ferrocenophanone [29], or of 4-hydroxybenzoyl-ferrocene and 4-hydroxypropio-phenone 7b [30], rearrangement to pinacolones occurred. Replacement of the ketone by an aldehyde, as in the reaction of 4-hydroxybenzaldehyde, 7d, with 6, led to an inseparable mixture of unknown products, and neither the chlorinated compound 3d nor the alcohol 5d was detected by mass spectrometry.
Alcohols 5a-c were obtained as mixtures of Z/E isomers. Even though the two forms could be separated by semi-preparative HPLC, the isomerization in solution did not allow the isolation of pure isomers except in the case of 5b for which the major isomer could be isolated from crystallization and identified as a Z isomer by 2D NMR. However, this isomer slowly isomerized in acetone to give an 86:14 Z/E mixture. In 5a and 5c, the observed Z/E ratios in acetone were 73/27 and 62/38, respectively. All reactions produced mixtures, and only the hydroxyl compounds could be isolated in pure form, albeit in relatively low yields.
Suitable crystals of the E isomer of 3-ferrocenyl-4-(4-hydroxyphenyl)-1-hydroxy-hex-3-ene, 5b, could be isolated for an X-ray crystallographic study. The molecular structure appears as Figure 1, and reveals that the ferrocenyl moiety is twisted out of the plane of the double bond though 30.5°, whereas the plane of the phenol ring is rotated through 68.6°, making a dihedral angle between the C 5 H 4 and aryl rings of 81.8°. The ethyl and hydroxyethyl substituents lie on opposite sides of the central plane and make a dihedral angle of 151.6°. Clearly, these orientations minimise steric interactions between the bulky substituents. The cyclopentadienyl rings in 5b are parallel and are separated by 3.274 Å. Interestingly, when a shorter chain compound such as chloroacetylferrocene 8 [31] was allowed to react with the mono-aryl ketone 7b, a different process occurred. As shown in Scheme 3, neither of the anticipated ethylenic compounds (chlorinated or hydroxylated) was formed, instead only the dehalogenated pinacolone 9 was isolated, in 29% yield. This compound corresponds to the formation of the more stable cation, situated α to the ferrocenyl moiety, with subsequent migration of the phenolic group onto the carbon bearing the positive charge. In principle, two pinacol-to-pinacolone rearrangement products are possible, but the migratory aptitude of the phenolic group is markedly higher than that of the alkyl substituent because of the greater ability of the former to delocalize positive charge in the transition state [32]. The identity of the pinacolone rearrangement product, 9, was determined not only spectro-scopically, but also by X-ray crystallography ( Figure 2), which established unequivocally that the phenolic group had migrated onto the carbon atom adjacent to the ferrocenyl substituent. The geometry at the central carbon, C(2), is almost perfectly tetrahedral with angles ranging only from 109.1° to 110.2°. The cyclopentadienyl rings in the ferrocenyl unit are parallel, and are separated by 3.331 Å; they make a dihedral angle of 85° with the plane of the phenol. In contrast, the reaction of chloroacetylferrocene, 8, and 4,4'-dihydroxybenzophenone, 7e, furnished the alkene 10 (49%), but no pinacol rearrangement product was observed (Scheme 3). Once again, however, the chlorine atom was replaced by a hydrogen atom, and the identity of 10 was confirmed by direct coupling of 7e with acetylferrocene, 11; in this latter case, 10 was obtained in 90% yield. The corresponding reaction between propionylferrocene, 12, and ketones 7b or 7f, yielded classical alkenes 4 or 13 in yields of 30% or 33%, respectively (Scheme 4). The rearrangement products 14 or 15 were also formed as contaminating by-products, and the X-ray crystal structure of 3-ferrocenyl-3-(4hydroxyphenyl)-hexan-4-one, 14, is shown as Figure 3.

O2
Having shown that alcohols 5a-c could be obtained in low yields from the chlorinated precursor 6, we sought a more rational route to these compounds. To improve the yield of 5b, we envisioned a McMurry cross-coupling reaction between 4-hydroxypropiophenone, 7b, and ethyl 3-ferrocenyl-3oxo-propanoate, 16 (Scheme 5), with the aim of reducing the ester 17a to the required alcohol 5b. However, McMurry coupling failed to give the desired ester 17a, the only material isolated being the homocoupling product, i.e., diethylstilbestrol (DES), together with recovered unreacted ketoester 16.

Scheme 5.
McMurry reaction between ketoester 16 and mono (7b) and diaryl (7e) ketones. In contrast, the reaction of the ketoester 16 and 4,4'-dihydroxypropiophenone, 7e, led to the alkene 17b, but in only 11% yield, and remaining ketoester was recovered. However, by using two equivalents of 4,4'-dihydroxybenzophenone, the cross-coupling yield could be increased to 40%, allowing the obtention of X-ray quality crystals of 17b, whose structure appears as Figure 4.  The C 5 H 4 ring of the ferrocenyl substituent is twisted through 27° out of the plane of the double bond, whereas the Z and E phenolic rings are rotated through 60.4° and 64.3°, respectively, in opposite directions, making a dihedral angle between them of 89.3°. The cyclopentadienyl rings in the ferrocenyl moiety are parallel and are separated by 3.329 Å. The results of all the coupling reactions are summarized in Table 1.
Alkene, and also isolation of a rearrangement product, 14 10 FcC Alkene when using 2 eq of (p-HOC 6 H 4 ) 2 C=O

Mechanistic Considerations
The failure of McMurry reactions of chloroalkyl-ferrocenyl ketones to yield chlorinated products raises interesting questions. In the case of chloroacetylferrocene, the chlorine was replaced by hydrogen (entries 1 and 2 of Table 1), whereas reactions of chloropropionylferrocene with mono-aryl ketones led to hydroxyethyl derivatives 5a-c (entries 4-6).
We consider first the case of chloroacetylferrocene, 8, originally prepared by Schlögl and Egger [31] by Friedel-Crafts acylation of ferrocene in the presence of AlCl 3 . In particular, we note their report that, when a solution of 8 in CS 2 or CH 2 Cl 2 was stirred with AlCl 3 for two hours at room temperature, it suffered dechlorination to form acetylferrocene, 11, whose identity was confirmed by independent synthesis. Apparently, abstraction of chloride by the Lewis acid to form AlCl 4 − leads initially to formation of the corresponding primary carbocation that is presumably stabilized as the strained three-membered cyclic ferrocenyl-stabilized cation 18 (Scheme 6). Reduction to form the methyl group in 11 requires formal addition of a hydride, and such a process is well-known to proceed under Friedel-Crafts conditions whereby abstraction of hydride can occur-even from an alkane! [33]. Now, assuming that a Lewis acidic titanium chloride can behave analogously by abstracting a chloride from 8, there is a straightforward explanation for the conversion of a chloroacetyl substituent into the corresponding acetyl derivative, as in 10 (entry 2). Turning now to the generation of the hydroxyethyl substituents in 5a-c, once again one can envisage abstraction of chloride to form a primary cation that cyclises to the somewhat less strained four-membered ferrocenyl-stabilized cation 19 (Scheme 7). Attachment of TiCl 2 to the two ketonic oxygens, addition of hydride and elimination of TiOCl 2 can lead to the observed products. However, the hydride abstraction process is rather inefficient and the observed yields are low. It is now apparent why hydroxyethylation is observed in reactions of chloropropionylferrocene, 6, with mono-aryl ketones when there is a possible source of hydride by abstraction from the alkyl group, but is not possible in reactions involving a diaryl ketone. This also provides a rationale for the non-formation of a hydroxyethyl derivative from 4-hydroxybenzaldehyde (entry 7 in Table 1).

Scheme 6.
Proposed mechanism for the Lewis acid promoted conversion of a chloromethyl substituent into a methyl group.

Biological Results
The IC 50 values (µM) of most of the new isolated compounds were measured on hormone-independent breast cancer cells MDA-MB-231, and are collected in Table 2. The IC 50 values of ferrocidiphenol 1 and ferrociphenol are also included for the sake of comparison [10,34]. It is noteworthy that compounds 5a-c, which contain the hydroxyethyl group, are more active than the related ferrocidiphenol, 1, with IC 50 values for 5a, 5b and 5c of 0.14, 0.28 and 0.36 µM, respectively. Indeed, compound 5a (IC 50 = 0.14 μM) is four times more potent than 1 (IC 50 = 0.60 μM [10]). We have shown previously that compounds of the ferrociphenol type have a higher activity when they possess two para phenol groups (or two biologically hydrolysable ester groups [36]) as compared to compounds bearing only one phenol group [34,37,38]. Hence, as listed in Table 2, the ferrocidiphenol, 1, with two phenol groups, is more active than its monophenol analogue, Fc(Et)C=CPh(C 6 H 4 OH), which shows an IC 50 value of 1.13 µM [34]. It appears, therefore, that the hydroxyethyl group imparts a stronger antiproliferative effect than a phenol group. Moreover, it appears that the length of the alkyl chain in the series 5a-c (methyl, ethyl and propyl) plays a significant role; even though the differences are not very large, it seems that the smaller the substituent the more active is the compound. The important role of the hydroxyethyl group is confirmed by the lower activity of compound 4 which lacks the hydroxyl function, whereby its IC 50 value (1.14 μM) is four times less active than 5b (IC 50 = 0.28 μM), and at the same level than the ferrociphenol with a single phenolic substituent. We note that replacement of the ethyl group of the ferrocidiphenol, 1, by a methyl group in compound 10 led to a decrease in activity. One might have anticipated that the smaller alkyl substituent would have led to somewhat enhanced activity, but the subtle change in electronic effects may have somewhat disfavored formation of the expected quinone methide metabolite [19].
Finally, we note that the pinacolone 9 is not very active, with an IC 50 value of only 3.13 µM. This again supports the idea that the [ferrocenyl-double bond-phenol] motif is crucial for antiproliferative activity in this series of compounds. Compound 9 has approximately the same IC 50 as ferrocenyl-di(4hydroxyphenyl)methane, 20, (3.5 µM) [35], which also possesses a saturated central carbon impairing electronic circulation.

General Information
All reactions and manipulations were carried out under an argon atmosphere using standard Schlenk techniques. THF was distilled over sodium/benzophenone prior to use. Thin layer chromatography was performed on silica gel 60 GF 254 . The semi-preparative HPLC separations were performed on a Shimadzu instrument with a Nucleodur C18 column (length of 25 cm, diameter of 3.2 cm, and particle size of 10 µm) using acetonitrile/water as eluent. IR spectra were obtained on a FT/IR-4100 JASCO 180 spectrometer. 1 H and 13 C-NMR spectra were acquired on a Bruker 300 MHz spectrometer, while 2D NMR spectra were obtained on a Bruker 400 MHz instrument. Mass spectrometry was carried out at the "Service de Spectrométrie de Masse" at ENSCP, Paris. High resolution mass spectra (HRMS) and RX structures were acquired in the "Institut Parisien de Chimie Moléculaire (IPCM-UMR 8232)" at the "Université Pierre et Marie Curie", Paris. Microanalyses were performed by the "Service de Microanalyse ICSN" at Gif sur Yvette, France. Determination of the cytotoxicity of the complexes was performed at IMAGIF (ICSN, Gif sur Yvette, France).

General Procedure for McMurry Coupling Reactions
Zinc powder was suspended in dry THF at room temperature in a Schlenk tube under argon, and titanium tetrachloride was added slowly via a syringe while stirring. The reaction mixture was heated at reflux for 2 h, after which time a THF solution containing the two ketones was added, and the mixture was heated at reflux for a further two hours. The reaction mixture was cooled, poured into water, acidified with HCl until the dark color disappeared and extracted with dichloromethane. The organic layer was washed with water, dried over magnesium sulfate, filtered and concentrated under reduced pressure. The crude mixture was chromatographed on silica gel column using a mixture of dichloromethane/acetone 90/10 as an eluent and/or purified by semi-preparative HPLC using a mixture of acetonitrile/water as an eluent. (4) (and rearrangement product 14). Zinc powder (1.82 g, 27.9 mmol), titanium tetrachloride (2.04 mL, 18.6 mmol), propionylferrocene (1.13 g, 4.65 mmol), p-hydroxypropiophenone (0.70 g, 4.65 mmol). After column chromatography followed by preparative HPLC (acetonitrile/water 90/10), and recrystallization from a pentane/diethylether, 4 was obtained as orange crystals (0.50 g, 30% yield) consisting of a mixture of Z and E isomers (63/37). 1 (14).      (9). Zinc powder (2.35 g, 36 mmol), titanium tetrachloride (2.64 mL, 24 mmol), chloroacetylferrocene (1.58 g, 6.0 mmol), p-hydroxypropiophenone (0.90 g, 6.0 mmol). After purification by HPLC (acetonitrile/water 80/20), and recrystallization from pentane/diethylether mixture, 9 was obtained as orange crystals (0.62 g, 29% yield). 1

X-ray Crystallography
Suitable crystals of 5b, 9, 14 and 17b were obtained from acetonitrile or acetone:hexane (for 5b). Crystallographic data are given in Table S1 in the Supporting Information. These have also been deposited at the Cambridge Crystallographic Data Centre: 5b (1007086), 9 (1007087) and 17b (1007088). Molecule 14 exhibited a minor crystallographic disorder giving rise to an apparent mirror plane so, although the structure is unambiguously established, we prefer not to quote precise bond lengths and angles. Unit cell data are listed in reference [39].

Conclusions
It has been shown here that the ferrocenyl-diaryl-ethylene series is not the only one active against hormone-independent breast cancer cells. The second p-hydroxyphenyl group can be replaced by an alkyl group, but in this case the presence of a hydroxyethyl group is necessary for good activity. In fact, the alkyl group seems not to play an essential role except for a decrease in activity as the chain lengthens. As a result of these factors, 3-ferrocenyl-4-(4-hydroxyphenyl)-1-hydroxy-pent-3-ene, 5a, bearing a methyl group adjacent to the phenol substituent, has the best value of the series (IC 50 = 0.14 µM). Unfortunately, at present the yields are low for the preparation of these promising hydroxylated species (10%-16% for the McMurry coupling), and need to be improved.
Even though one starts from relatively inexpensive, commercially available ketones and easily synthesized chloropropionyl-ferrocene, 6, (Friedel-Crafts), and the McMurry second step led to the cross-coupled product possessing the double bond and the hydroxyethyl substituent in a one-pot process, nevertheless, final separation of the desired product is difficult. Modified reaction conditions, together with an enhanced understanding of the mechanism (such as the importance of hydride transfer) may lead to improved yields. It may be worthwhile to attempt the McMurry reaction starting directly from the keto-alcohol, but pinacolone by-products would of course diminish the overall yields, and alternative routes need to be developed. Other new compounds bearing the [ferrocenyl-(hydroxyethyl)-ene-phenol] motif are under active investigation with the aim of elucidating the role of the hydroxyethyl group in promoting cytotoxic activity.