Synthesis and receptor binding in trans-CD ring-fused A-CD estrogens: Comparison with the cis-fused isomers

Abstract

Ligands which selectively activate only one of the estrogen receptors, ERα or ERβ, are current pharmaceutical targets. Previously, we have reported on substituted cis A-CD ligands in which the B-ring of the steroidal structure has been removed and cis refers the stereochemistry of the CD ring junction as compared to trans in estradiol. These compounds often showed good potency and selectivity for ERβ. Here we report the synthesis and binding affinities for a similar series of trans A-CD ligands, and compare them to the cis-series. Counterintuitively, trans A-CD ligands, which are structurally more closely related to the natural ligand estradiol, show weaker binding and less β-selectivity than their cis-counterparts.

Current studies of the mechanism of disease initiation in breast cancer and several other cancers have focussed on the pathway starting from the natural estrogen 17β-estradiol (E2), which by P450 aromatic hydroxylation can form the catechol estrogen and potentially the corresponding ortho-quinones [1]. Of various possible metabolites, the 3,4-quinone has been shown to intercalate into DNA and undergo Michael addition with DNA bases, ultimately forming depurinating adducts. DNA repair then leads to a high mutation rate and tumor initiation [2]. Once initiated, activation of the estrogen receptor ERα by endogenous ligands such as E2 has been demonstrated to promote proliferation of the cancer cells [3]. On the other hand, in some experiments activation of the receptor ERβ may suppress cancer cell growth [4], hence the current intense interest in developing ERβ-selective ligands [5].

We have recently reported the synthesis of a series of estrogen agonists represented by structure 1 shown in Fig. 1 in which the A ring and C rings are connected by a single bond [6–8], which we called A-CD estrogens. These compounds, especially those having electronegative substituents at C5, showed high affinity binding to the estrogen receptors (ERs), which for ERβ was in some cases even stronger than the naturally occurring ligand E2, which has the highest affinity of the natural ligands. We also observed some, and at times considerable, enhancement of ERβ- receptor binding selectivity. In general, the compounds which showed this type of binding selectivity also acted as strong β-selective agonists; such compounds are denoted “super-agonists” when they exceed the maximum level of transcription activation by E2 [9]. In our initial report [6], we presented these compounds as having the general structure 2 (see Fig. 1), where the CD ring junction is transas in E2. Subsequently [7,8], we corrected the structure assignment and showed that the compounds we had prepared actually had the cis-CD ring structure, as shown below in the parent A-CD compound for series 1.

Figure 1.

Stereochemical relationship between the natural estrogen E2, the A-CD estrogens, cis-series 1 and the trans-series 2. The cis and trans designations refer to the stereochemistry at the CD ring junction.

We now present our work on the preparation of a series of compounds of structural type 2 which have the trans-CD ring junction. We compare their binding affinity to both estrogen receptors with the corresponding cis-compounds in series 1. One might expect that retaining the geometry of the natural ligand at the CD interface would lead to optimum binding of the A-CD system. However, we will show that the unintended synthesis of compounds with structure 1 rather than 2 turns out to have been quite fortuitous, since the A-CD compounds that most closely resemble E2 generally bind less strongly and with lower selectivity to ERβ than those in the cis-series 1. As before, for convenience we have chosen the numbering in these compounds to correspond to the same numbering scheme as in E2 (see Fig. 1)

The trans-CD ring junction in the structures for series 2 was installed in one of two ways [10]. In the first approach (Scheme 2), the CD ring ketone 4 having the trans-ring junction was prepared following the method of Hajos and Parrish [11]. This ketone was then reacted with the lithio-derivative of a suitably protected ring A (moiety 5) to give the alcohol 6. Dehydration, with concomitant removal of the 3-phenol and 17-alcohol protecting groups, generally MEM-ethers, gave exclusively the alkene 7. The typical overall yield over these two steps was in the 80–90% range. Simple hydrogenation with Pd/C as catalyst afforded quantitatively a mixture of isomers at C9 from which the major desired isomer 2 was isolated by preparative HPLC. The structure of the parent compound in the trans-series was verified by an X-Ray structure determination, shown in Fig. 2 for 2a, the trans-CD parent compound.

Scheme 2.

Figure 2.

X-ray structure determination of compound 2a, the trans A-CD parent compound.

The ¹H NMR spectrum of 2a showed the C13 methyl group at 0.83 ppm; this compared with 1.05 ppm for the cis-CD-fused derivative 1. This difference was consistent for all other pairs of compounds described herein. ¹³C NMR also showed significant differences for the C17, which helped subsequently to verify whether either the trans- or the cis- CD ring juncture was present in derivatives of 1 and 2.

Alternatively, reaction of 5 with the MEM-protected enone 8 yielded the allylic alcohol 9. Acid-catalyzed dehydration with concomitant deprotection afforded the diene 9 (Scheme 2) in excellent yield. Hydrogenation with a variety of common catalysts, including Pd/C, gave mainly the compound 2, as shown by ¹H NMR examination of the crude reaction product. Small amounts of the three other possible stereoisomers were also formed. The formation of 2 as the major product in the hydrogenation of 10 indicates that the C14–C15 double bond is hydrogenated first, with preferential delivery of hydrogen from the less hindered side, to give the C8–C11 alkene 11 having the trans-CD ring junction. Subsequent hydrogenation gives, as expected, mainly the isomer 2 with the (S) configuration at C9. In contrast, hydrogenation of the cis CD-junctioned alkene 12 gives a close to a 1:1 mixture of stereoisomers at C9.

Either of the above sequences or slight variations thereof were used to prepare a series of 10 derivatives carrying a variety of substituents in the A-ring [10]. [For example, the 5-hydroxy derivative 2i was obtained in three steps by first converting 1,3-dibenzyloxy-4-bromobenzene into its lithio-derivative and reaction of this species with the enone 8. Purification of the reaction mixture via silica gel chromatography (Scheme 3) resulted in formation of the diene 1, which was hydrogenated to give a mixture of compounds from which 2i was obtained after reverse phase HPLC purification.

Scheme 3.

As pointed out above, the isomeric compounds 1 and 2 were most readily distinguished by focusing on the quaternary methyl group. In all of the compounds described these methyl hydrogens absorbed in the 0.83–0.85 ppm range for the trans-compounds but more downfield at 1.09 to 1.11 ppm for the cis-isomers. In the ¹³C spectrum this carbon was found near 12 ppm in the trans-compounds but again considerably more downfield near 20 ppm for the corresponding cis-isomers [10,11].

Binding affinities to the estrogen receptors ERα and ERβ are given as relative binding affinities (RBAs), where the RBA for estradiol on both ERs is 100. The RBA value and selectivity ratio RBA(ERβ)/RBA(ERα) for a series of 10 cis A-CD (series 1) and 10 trans A-CD compounds (series 2) containing saturated C-rings are given in Table 1. The binding affinity ratios RBA(trans)/RBA(cis) for both receptors are given in Table 2.

Table 1.

Relative binding affinities for cis-series 1 and trans-series 2 to estrogen receptors ERα and ERβ, selectivity ratio for ERβ/ERα. Error bars in parentheses.

cis A-CD structures (1)				trans A-CD structures (2)
Compound	RBA(α)	RBA(β)	β/α	RBA(α)	RBA(β)	β/α)
a (H)	1.47(0.26)	21.5(4.6)	14.6	2.38(0.19)	10(1.3)	4.2
b(4-F)	1.04(0.09)	8.7(1.5)	8.4	1.68(0.15)	6.84(0.41)	4.1
c(5-F)	27.3(0.70)	135.5(7.3)	5.0	4.22(0.06)	13.6(0.35)	3.2
d(5-Me)	2.82(0.45)	33.6(6.2)	11.9	0.47(0.1)	1.3(0.3)	2.8
e(5-CF3)	89.7(13.8)	205(23)	2.3	4.8(1.1)	4.9(0.1)	1.0
f(4,5-diF)	4.62(0.93)	42.8(5.5)	9.3	0.92(0.15)	7.3(1.8)	7.9
g(2,4,5-triF)	0.186(0.01)	1.73(0.02)	9.3	0.029(0.004)	0.097(0.025)	3.3
h (2,4-diF 5-Me)	0.75(0.20)	5.35(0.88)	7.1	0.08(0.007)	0.226(0.04)	2.8
i(5-OH)	0.26(0.07)	3.97(0.08)	15.3	0.037(0.003)	0.15(0.04)	4.1
j(5-OMe)	0.016(0.002)	0.123(0.03)	7.7	0.010(0.001)	0.037(0.008)	3.7
Estradiol (E2)				100.	100.	1.0
Average β/α ratio*			9.1			3.7

^*E2 is not included in the averages.

Table 2.

Ratio of binding to ERα and ERβ for the trans-compounds 2 vs the cis-compounds 1.

Compound	ERα	ERβ
a(Parent, R=H)	1.62	0.47
b(4-F)	1.58	0.79
c(5-F)	0.15	0.10
d(5-Me)	0.17	0.04
e(5-CF3)	0.054	0.024
f(4,5-diF)	0.20	0.17
g(2,4,5-triF)	0.16	0.056
h(2,4-diF 5-Me)	0.11	0.042
i(5-OH)	0.14	0.038
j(5-OMe	0.63	0.30
Average β/α ratio	0.48	0.20

The values of the RBAs shown in Table 1 are correlated via a sigmoidal dependence with the transcription activation, or “potency”, of each ligand in its receptor [8], so the RBAs are already useful predictors of potential drug activity. With four distinct data sets (cis- and trans-ligands in two receptors, ERα and ERβ) there are a number of interesting comparisons that can be made. First, consider the selectivity of trans vs. cis ligands binding to ERα and ERβ. The β-selectivity of the parent trans compound 2a, for example, is given by the β/α ratio = 10/2.38= 4.2 (see Table 1), whereas for the cis compound 1a it is 21.5/1.47 = 14.6, over three times larger. Looking at the selectivity ratios β/α for each series, the average value for the ligands in the trans-series is 3.7, whereas for the cis-series it is 9.1. Thus, the cis-series is much more β-selective than the trans-series.

It is also of interest to compare the magnitude of the RBAs of the trans vs. the cis compounds in Table 1. For example, comparing the trans/cis RBA ratio for compound c (2c/1c) in ERα, the ratio is 4.22/27.3 = 0.15 (see Table 2); the cis-structure is much more strongly bound. Continuing in this way, the average ratio for trans/cis binding into ERα for the 10 ligands is 0.48, showing that the binding of the cis-compounds into ERα is, generally speaking, over twice as strong as that of the trans-compounds. Performing the same operations on binding to the ERβ receptor, the trans/cis ratio for compound a (2a/1a) is 10/21.5 = 0.47. The average for the whole series is only 0.20, so binding to the ERβ receptor favors the cis-structures by a factor of five.

It should also be pointed out that in the cis series 1, the compounds such as 1c (5-F) and 1e (5-CF3), which showed the highest binding affinities, also had the lowest β/α selectivity, 5.0 and 2.3, respectively. In contrast, some ligands with more modest binding affinities to both receptors 1a (parent) and 1i (5-OH) showed the maximum selectivity for the ERβ to ERα ratio of ca. 15. The very low binding affinity of the 5-OCH₃ derivatives 1j relative to those carrying CF₃, CH₃ and OH groups at this position was unexpected. For comparison, the binding affinity to the ERβ of 1e (5-CF₃) is about 1700 times greater than that of 1j. A similar comparison for 2e vs 2j gives a value of close to 130. It should be possible to explain these large differences by examining the interaction of these substituents with receptor residues, work that we have already begun.

Since the results presented above were so unexpected, we looked at two more series of compounds with the trans-, cis-CD ring fusion (Series 7 and Series 14, Table 3). These were prepared during the course of our studies and are C9–C11 alkenes having the trans CD ring fusion (Series 7) and the cis fusion (Series 14). The compounds in Series 7 were obtained as intermediates in the synthesis via Scheme 2; those in Series 14 were formed in the preparation of the cis A-CD compounds [10]. As shown in Table 3, even though relatively few pairs are currently available, the trends discussed above are also seen for these sets of isomers. Thus, the average β/α selectivity ratio for the cis-series 7 (6.6) is higher than that for the trans-series 14 (2.7) by a factor of 2.4; this is identical to the ratio for the cis-series 1 and trans-series 2 (factor of 2.4). Since the alkene series show no special advantages over the alkanes and is more likely to be reactive in vivo, and since the cis-series 1 shows improved β-selectivity and binding affinity compared to the trans-series 2, future drug development using these compounds will be focussed on the cis-series 1.

Table 3.

Relative binding affinities for cis-series 14 and trans-series 7 to estrogen receptors ERα and ERβ, and selectivity ratio for ERβ/ERα.

cis A-CD structures (14)				trans A-CD structures (7)
Compound	RBA(α)	RBA(β)	β/α	RBA(α)	RBA(β)	β/α)
a(H)	0.39(0.04)	6.8(0.6)	17	2.38(0.19)	17.3(2.4)	4.7
b(5-F)	4.5*(0.8)	49*(7)	11*	21.7(0.4)	38(11)	1.7
c(5-Cl)	60(8)	118(4)	2.0	9.7(2.2)	14.1(3.8)	1.5
d(5-CF3)	122(11)	174(8)	1.4	------	------	------
e(5-OMe)	0.004(0.001)	0.022(0.004)	5.5	0.006(0.001)	0.029(0.002)	4.8
f(2,4,5-triF)	------	-------	------	0.099(0.33)	0.076(0.8)	0.8
g (2,5-diF)	0.44(0.02)	2.7(0.5)	6.1
H(4-F,5-Cl)	50(6)	149(9)	3.0
I(4,5-diCl))	3.0(0.8)	55(9)	18.3
j(2,4-diF 5-Cl)	3.35(0.06)	6(2)	1.8
	Average β/α ratio		6.6			2.7

^*Determined for a ~1:1 mixture of 8–9 and 9–11 alkenes

We have begun a series of modeling studies aimed at understanding the trends in reactivity described above [8,14]. In future work we will focus on understanding the origins of the ERβ-selectivity described in this paper, as well as the preference for the cis- rather than the trans-geometry for the binding affinity in the receptors ERα and ERβ.

Scheme 1.

Abbreviations

RBA

relative binding affinity

ER

estrogen receptor

E2

17β-estradiol