C3 Halogen and C8’’ Substituents on Stilbene Arotinoids Modulate Retinoic Acid Receptor Subtype Function
Introduction
Retinoic acid receptors (RARs; subtypes a, b, and g)[1] are mem- bers of the nuclear receptor (NR) superfamily, which upon ligand activation act as transcription factors that bind DNA reg- ulatory elements in the promoter regions of target genes.[2] There is compelling evidence that retinoid signaling in both embryos and adults is mediated by RAR–RXR heterodimers.[3,4] Other members of the NR superfamily, including VDR, PPAR, and TR operate by the same mechanism of heterodimer formation with the promiscuous retinoid X receptor (RXR) as partner;the ligand binding domains of NRs. The ligand-induced struc- tural alterations modify the co-repressor surface, resulting in CoR dissociation. Concomitantly an overlapping but distinct novel surface is created which can establish an interface with the “NR boxes” of co-activators (CoAs) such as the prototypic p160/SRC proteins.[16] These co-activators, in turn, recruit his- tone acetyltransferases (HATs, such as CBP and p300), which have the opposite enzymatic activity of HDACs and de-repress some of those dimers play a major role in the etiology and therapy of metabolic diseases.[5,6]
In the absence of the cognate ligand, RAR–RXR heterodim- ers are nuclear, bind to the DNA response elements of (some) target genes and recruit co-repressors, such as NCoR or SMRT, which in turn bind transcription repressive factors such as his- tone deacetylases (HDACs). The ultimate outcome is the estab- lishment of “co-repressor complexes”, which silence the corre- sponding target genes through chromatin condensation upon histone deacetylation.[7–10] The structural basis of co-repressor binding to apo-NRs by virtue of their “CoRNR boxes” is rather well understood from both functional[11–13] and crystallograph- ic[14] studies with antagonists, some of which can induce or even enforce[15] co-repressor interaction. Agonist (that is, the natural hormone) binding induces a major allosteric effect in chromatin-mediated silencing. A variety of other factors and machineries are recruited to (or dissociated from) the NR bind- ing site by this allosteric switch, possibly in a highly dynamic manner.[17] Conceptually, an agonist is thus a transcription-acti- vating ligand because it induces allosteric conformational changes that allow the establishment of co-activator com- plexes at the chromatinated target site. Hence, the NR inter- prets the inherent structural information of the agonist and converts it into chromatin struc- ture-modulating events relevant for transcription regulation. In- terestingly, these allosteric transi- tions can be variously modulat- ed by ligand design, generating NRs with very different types of interaction patterns with co-reg- ulatory factors.[15] Neutral antag- onists also induce dissociation of CoRs, but also decrease the af- finity of CoA interaction below the levels observed with the unsystematically modify the ligand structure at critical positions to induce divergent or receptor subtype-selective allosteric ef- fects that confer a particular type of agonistic or antagonistic functionality onto the targeted receptor.
Preliminary characterization[34] of a collection of arotinoids[35] with a stilbenoid structure based on the pan-agonist TTNPB[35] (3, Figure 1) led to the puzzling and intriguing observations context of the RAR–RXR heterodimer.[15] Structural and func- tional studies demonstrate that the anchoring of H12 is funda- mentally distinct in the presence of antagonists and inverse agonists.[15] Note also that in the case of the RAR–RXR hetero- dimer, binding of a rexinoid agonist to RXR also induces a CoA binding surface at the RXR subunit, but this does not result in dissociation of the CoR from apo-RAR, thus precluding CoA re- cruitment and accounting for what has been called “RXR sub- ordination”.[19,20]
The ability to regulate gene networks that control cell growth, differentiation, survival, and death makes retinoic acid receptors key players of multiple physiological processes, in- cluding embryonic development and organ homeostasis.[3,21,22] In particular, several lines of evidence support the requirement of the RARb subtype for the antiproliferative effect exerted by retinoids.[23–31] Therefore, RARb-selective modulators[23] are in- teresting pharmacological tools to decipher the antitumor ac- tivities of retinoids and pave the way toward refined retinoid- based treatment paradigms.
The synthesis of retinoids with improved receptor and/or functional selectivity is complicated by both the highly con- served nature of the amino acids surrounding the ligand in the ligand binding pocket (LBP) of retinoic acid receptors[32] and the observation that agonists for one receptor can be antago- nists or even inverse agonists for another subtype. Indeed, RARb and RARa differ in only one residue, whereas RARb and RARg differ in two, and these differences are known to deter- mine subtype selectivity. Therefore, even though we have been able to establish general guidelines for the design of sub- type-selective retinoids,[33] additional systematic structure–func- tion studies are necessary to uncover the functional conse- quences of ligand modification. It is particularly important to that some retinoids can function as potent antagonists of one or two subtypes and agonists of the other. An example is illus- trated by BMS453 (4 a), a potent antagonist of TTNPB-induced transcription for RARa and RARg and a ’mixed’ agonist/antago- nist (that is, a weaker agonist than TTNPB 3) for RARb.[36] An at- tractive hypothesis is that differently restricted or accessible re- gions of the RAR ligand binding pocket account for selective binding of ligands. However, structural evidence to validate this hypothesis is still insufficient, as only the crystal structure of an antagonist, BMS614 (2), bound to RARa has been report- ed.[37]
Interestingly, BMS641 (4 c), a derivative of BMS453 (4 a) with a chlorine atom at position C3, displayed an overall decrease in RAR affinity which is more pronounced for RARa, and there- fore RARb selectivity is indirectly improved [(Kd = 2.5 nM); cf. RARa (Kd = 225 nM) or RARg (Kd = 223 nM)].[36] A similar effect of the halogen atom was observed on the activities of TTNPB derivatives 3-Cl-TTNPB (3 b) and 3-Br-TTNPB (3 c).[36]
Results
Chemistry
The synthesis started with the construction of known 7- bromo-4,4-dimethyl-3,4-dihydronaphthalen-1(2H)-one 10.[38,39] Treatment of ethyl 4-oxopentanoate 7 with methyl magnesium bromide and ensuing lactonization provided 8 in 67 % yield. Friedel–Crafts acylation/intramolecular alkylation of benzene with lactone 8 was effected with aluminum trichloride and generated tetralone 9 in 73 % yield. Bromination of 9 with Br2 and catalytic quantities of aluminum trichloride provided ex- clusively 10[39] in 77 % yield (Scheme 1).
Addition of the commercially available Grignard reagents de- rived from bromobenzene and p-tolyl bromide to a solution of 10 in tetrahydrofuran (THF) at ambient temperature afforded benzyl alcohols 11 and 12[39] in 92 and 65 % yield, respectively. In the same manner, addition to 10 of the organolithium pre- pared in situ by reaction of phenylacetylene with n-butyllithi- um in THF provided 1-ethynyl derivative 13 (82 %). Alcohols 11–13 were converted into the dihydronaphthalenes 14–16 by heating at reflux in benzene with catalytic quantities of p-tol- uenesulfonic acid. Transformation of aryl bromides 14–16 into the corresponding aldehydes 17–19 was effected by halogen– metal exchange using tert-butyllithium at —78 8C, and trapping the aryllithium intermediate with N,N-dimethylformamide (77, 76, and 79 %, respectively for 17, 18, and 19).
The required phosphonates were prepared as illustrated in Scheme 1, starting from the corresponding commercial m-halo- genated p-methylbenzoic acids 20 b–e. These were converted into the methyl esters 21 b–e in high yields (91–97 %) upon treatment with trimethylsilyldiazomethane and methanol in benzene, and the esters were brominated at the benzylic posi- tion under radical conditions, with Br2 in carbon tetrachloride by irradiation with a tungsten lamp (250 W). Treatment of the benzyl bromides 22 b–e with triethyl phosphite at 150 8C under classical Arbuzov conditions afforded the corresponding phosphonates 23 b–e in high yields (79–94 %).
Transcription activation studies
To evaluate the effects of the described retinoids on RARa-, RARb-, RARg-, and RXRb-mediated transactivation, a reporter assay with genetically engineered HeLa cell lines[34] was used. Because the amino acid residues that constitute the ligand binding pockets of the RXR subtypes do not differ, a single RXR reporter cell line (expressing RXRb LBD) is assumed to reveal the ligand responsiveness
as readout for all three RXRs. The reporter cell lines are engi- neered to express a fusion pro- tein, comprising the ligand bind- ing domain of the correspond- ing receptor and the DNA bind- ing domain of the yeast Gal4 transcription factor. In addition, the cells contain a stably inte- grated luciferase reporter gene, which is controlled by five Gal4 response elements in front of a b-globin promoter; this is termed “(17m)5-bG-Luc”.[19,41] The transcriptional activity of the var- ious compounds was compared with the activity of the pan-RAR agonist TTNPB (3 a) and 9-cis-ret- inoic acid as positive controls for RAR and RXR, respectively. Note that this reporter system is in- sensitive to endogenous recep- tors, which cannot recognize the Gal4 binding site.
Some observation within the series 4 and 6 are particularly important with respect to RARb. While the introduction of fluorine in 4a increases its agonis- tic activity, the antagonistic activity relative to TTNPB is also augmented. However, switching to chlorine or bromine completely obliterates this activity such that both 4c and 4d are potent RARb agonists without RARg activity, while some RARa antagonist activity is still present. Unexpectedly, replacing bromine by iodine reinforces/re-establishes the antagonistic activ- ity for all three RAR subtypes. Two other structure– function correlations were unexpected: One concerns the ability of bromine to generate a pure RARb ago- nist (compound 6 d) in series 6 in which analogues 6 a–c are RARb inverse agonists and 6e is an antago- nist; the other concerns the swapping of agonist/ antagonist characteristics in 6b to 6e when acting as ligands for RARg.
Taken together, the above data confirm the critical function of C8’’ substitution and C3 halogenation in stilbene arotinoids, which together dictate the func- tional outcome of a ligand as agonist, partial agonist/ antagonist, pure antagonist, or inverse agonist for a given RAR receptor subtype. As anticipated from ligand design, similar to the parent arotinoids 4 a, 5 a, and 6 a, the halogenated derivatives were inactive with RXRb (Supporting Information) and therefore acted as pure ligands for the RAR subtypes.
We attempted to interpret the activity profiles of the stilbene arotinoids reported herein by molecular modeling based on currently available structural in- formation. Given that no crystal structure is yet available for retinoid inverse agonists, the analysis was restricted to the modeling of ligand–LBP complexes of agonists and antag- onists, for which several structures have been reported,[36,42–46] among them the agonist-bound RARb[36] and the antagonist- bound[37] RARa LBPs. We therefore chose to model compound 4 b, which shows the contrasting selectivity of being an RARb agonist and an RARa antagonist. Ligand 4b was positioned in the active site on the basis of the crystallographic structures of acquisition of agonist activity for RARb and RARg. These data support the view that there is a delicate balance between the nature of the C3 halogens and the bulkiness/length of the C8’’ substituents which is likely due to allosteric effects on the abili- ty of the receptor LBP to interact with co-regulators or to sta- bilize or weaken such interactions. In keeping with this notion, within each series C3 halogenation can have very selective ef- fects. In particular: 4) Increasing halogen atom size frequently the complexes of RAR with several ligands.[36,42–46] Preferred docking sites for functional groups were evaluated with the program GRID,[47] which assisted in the selection of binding modes.
The resulting structure of 4b bound to RARb (Figure 5) has the phenyl substituent at C8’’ oriented toward H5, and this
conformation would explain the discriminating agonism toward RARb.[36] The enlarged ligand binding pocket of RARb (~ 16 % greater for the TTNPB–RARb complex relative to that of the 9-cis-retinoic acid–RARg complex; see the Supporting Infor- mation) allows the occupation of the additional cavity between H5 and H10 by the phenyl substituent at C8’’ of 4 b.
The fluorine atom was found facing Ala 225 at short van der Waals contacts. The fluorine moiety also has close contact with the backbone carbonyl group of Phe 221, which probably cor- responds to a halogen bond (C—X···O—Y, for which C—X is a carbon-bonded halogen and O—Y is a carbonyl, hydroxy, charged carboxylate, or phosphate group)[48] with the appro- priate geometry that contributes to the stabilization of the complex. This halogen is considered to also contribute to the selectivity, as it allows differentiation with RARa due to the steric hindrance resulting from the presence of a serine residue (Ser 232) in H3 of RARa instead of alanines in RARb/g (Ala 225 in RARb).
Ligand 4b complexed to the RARa antagonist-bound form was placed in the same conformation as before, but the phenyl group was oriented toward H12, and the fluorine moiety was found closer to the backbone carbonyl group of Leu 269. Relative to the RARb complex, 4b was found flipped over the longitudinal axis (Figure 5). The more restricted bind- ing pocket forces a conformational adaptation of the ligand in its interaction with RARa.
Discussion
As with most biological phenomena, the mechanism of gene transcription mediated by nuclear receptors consists of a com- plex yet orchestrated and tightly regulated sequence of signal- ing events leading to gene repression or gene activation. The detailed description of this process is slowly emerging, primari- ly from crystallographic studies of the complexes between the ligand-bound and unbound domains of the NRs and their co- regulatory effectors,[49] and functional studies of the changes in chromatin structure that modulate downstream events relevant for transcriptional regulation.[50–52]
The RARb subtype is required for the antiproliferative effect of retinoids,[23] as the expression of RARb is epigenetically si- lenced in breast cancer[30,31] or frequently lost in many neoplas- tic tissues.[24–28] Moreover, the reactivation of RARb expression after retinoic acid treatment has been associated with a favora- ble clinical response in some cases.[29]
RARb-selective agonists have been described,[23] including a potent and selective agonist of the RARb2 isoform (one of the four isoforms originated by differential use of two promoters and alternative splicing)[1] with druglike properties and good oral bioavailability in rats.[53,54] RARb subtype-selective antago- nists and inverse agonists are, however, unknown.
Based on a preliminary screen, we focused on the stilbene arotinoids related to the potent pan-agonist TTNPB (3)[36] for development of subtype-selective antagonists. The underlying rationale for the present study was to systematically modify two positions on the parent non-halogenated arotinoid BMS453 (4 a), which displays a mixed agonist/antagonist activi- ty with the three RAR subtypes (Table 1). These two substitutions concern the C8’’ position, which is critical for the generation of inverse agonistic activity,[15] and C3, at which halogena- tion contributes to RARb selectivity.[36] We have synthesized and evaluated the entire collection of analogues that differ in the nature of the halogen atoms (b–e: F–I) and in the bulki- ness of the C8’’ group (phenyl 4, p-tolyl 5, and phenylethynyl 6).
Conclusions
Stilbene arotinoids with C3 halogens and C8’’-phenyl, p-tolyl, and phenylethynyl groups were synthesized. The functional effects of both substituents to account for the role of the ligand as agonist/antagonist/inverse agonist of RAR subtypes. In particular, 4a and 4b are potent agonists of the RARb sub- type and antagonists of RARa. Several inverse agonists with selectivity for RARa/b were also characterized upon increasing the bulkiness of the groups at C8’’.
A combination of steric and electronic factors of the ligand substituents acting in various regions of the LBP dictates the conformation of the stilbene single bonds upon interaction with the residues of the LBP. In limiting cases, the ligand might even flip over its longitudinal axis, forcing the groups at C8’’ to occupy alternative regions of the LBP of the RAR subtypes, thus explaining the divergent activities.
Conceptually the agonist/antagonist profiles exhibited by these analogues with the various subtypes are likely to be un- coupled from the binding event itself and are related to the subtype-specific allosteric effects that generate novel or alter pre-existing surfaces that mediate the interaction with co-regu- lator complexes.[15]
The SAR studies reported herein for the stilbene class of aro- tinoids provides significant advances on our understanding of the modulation of RARb with small molecules. Whereas the mechanism of the tumor-suppressing role of RARb still remains elusive, the possibility to selectively modulate its activity in a conditional manner through the use of ligands promises to reveal important clues. In this regard, the lack of selective and powerful antagonists is a limitation, but our research groups are actively engaged in this endeavor.
Two-hybrid assays. Gal-SMRT and Gal-NCoR were described previ- ously.[36] VP16-hRARa(DEF) and VP16-hRARb(DEF) plasmids corre- spond to a fusion between the activation domain VP16 and resi- dues 153–448 of hRARa, and residues 146–462 of hRARb, respec- tively. For two-hybrid assays, 1.2 × 105 HeLa cells were seeded in 24-well plates. After attachment, cells were co-transfected with 50 ng Gal4 chimera, 40 ng VP16 chimera, 200 ng (17m)5-bG-Luc and 50 ng CMV-bGal, according to the Jet-Pei (Polyplus transfection) protocol. After 18 h, cells were treated with vehicle or ligand at 1 mM. After a further 16 h, cells were lysed in 1 × passive lysis buffer (Promega). Reporter gene activity was analyzed by standard protocol for luciferase assay and quantified with a luminometer (Berthold Technologies). The results were normalized to the b-galactosidase BMS493 activities.