Modeling, Synthesis and Biological Evaluation of Potential RetinoidX Receptor-Selective Agonists: Novel Halogenated Analogues of 4-[1-(3,5,5,8,8-Pentamethyl-5,6,7,8-tetrahydro-2-naphthyl)ethynyl]benzoic Acid (Bexarotene)

MED

DOI: 10.1002/cmdc.201200319

Modeling, Synthesis and Biological Evaluation of Potential

Retinoid X Receptor-Selective Agonists: Novel

Halogenated Analogues of 4-[1-(3,5,5,8,8-Pentamethyl-

5,6,7,8-tetrahydro-2-naphthyl)ethynyl]benzoic Acid

(Bexarotene)

Julie K. Furmick,^[a]Ichiro Kaneko,^[a]Angela N. Walsh,^[a]Joanna Yang,^[a]Jaskaran S. Bhogal,^[a]

Geoffrey M. Gray,^[c]Juan C. Baso,^[c]Drew O. Browder,^[a]Jessica L. S. Prentice,^[a]

Luis A. Montano,^[a]Chanh C. Huynh,^[a]Lisa M. Marcus,^[a]Dorian G. Tsosie,^[a]

Jungeun S. Kwon,^[a]Alexis Quezada,^[a]Nicole M. Reyes,^[a]Brittney Lemming,^[a]Puneet Saini,^[a]

Arjan van der Vaart,^[c]Thomas L. Groy,^[b]Pamela A. Marshall,^[a]Peter W. Jurutka,^[a]and

Carl E. Wagner*^[a]

The synthesis of halogenated analogues of 4-[1-(3,5,5,8,8-pen-

tamethyl-5,6,7,8-tetrahydro-2-naphthyl)ethynyl]benzoic acid (1),

known commonly as bexarotene, and their evaluation for reti-

noid X receptor (RXR)-specific agonist performance is de-

scribed. Compound 1 is FDA approved to treat cutaneous T-

cell lymphoma (CTCL); however, bexarotene treatment can

induce hypothyroidism and elevated triglyceride levels, pre-

sumably by disrupting RXR heterodimer pathways for other

nuclear receptors. The novel halogenated analogues in this

study were modeled and assessed for their ability to bind to

RXR and stimulate RXR homodimerization in an RXRE-mediated

transcriptional assay as well as an RXR mammalian-2-hybrid

assay. In an array of eight novel compounds, four analogues

were discovered to promote RXR-mediated transcription with

EC₅₀values similar to that of 1 and are selective RXR agonists.

Our approach also uncovered a periodic trend of increased

binding and homodimerization of RXR when substituting a hal-

ogen atom for a proton ortho to the carboxylic acid on 1.

Introduction

Retinoids are small molecules that interact with at least two

specific nuclear receptors and regulate cellular processes such

as gene transcription, proliferation and differentiation. The two

major nuclear receptors that retinoids target are retinoid X re-

ceptors (RXRs) and retinoic acid receptors (RARs), and both re-

ceptors have three known subtypes: a, b and g.^[1]Retinoid re-

ceptors, in addition to other lipophilic hormone molecule re-

ceptors, are part of a larger receptor superfamily, including the

vitamin D receptor (VDR) and thyroid hormone receptor (TR),

all of which essentially function to promote transcription in the

presence of an appropriate molecular signal. This molecular

signal, usually comprised of an endogenous ligand, binds in

the protein’s ligand binding pocket (LBP), inducing a conforma-

tional change that ultimately enables the protein to bind to

a hormone responsive element (HRE) specific for the receptor

on DNA. Many HREs are located in or proximal to the promoter

regions for the genes they regulate, although a growing

number of these elements can be found at large distances up-

stream or downstream of the gene they control. Nonetheless,

HREs are typically constructed of minimal core hexad sequen-

ces consisting of half-sites interrupted by nucleotide spacers of

variable length between inverted, everted or direct repeats.^[2]

Nuclear receptors activate transcription by binding to the HREs

as heterodimers or homodimers, where each partner binds to

a half-site in the element.

Although initially proposed to operate as homodimers,^[3]

HRE high affinity binding by RAR, VDR and TR actually pro-

ceeds via an RXR heterodimer.^[4]When RXR binds to its natural

9-cis retinoic acid (9-cis-RA) ligand, it forms a homodimer that

associates with the RXR responsive element or RXRE, but when

functioning as a heterodimeric partner with other receptors,

[a] J. K. Furmick, Dr. I. Kaneko, A. N. Walsh, J. Yang, J. S. Bhogal, D. O. Browder,

J. L. S. Prentice, L. A. Montano, C. C. Huynh, L. M. Marcus, D. G. Tsosie,

J. S. Kwon, A. Quezada, N. M. Reyes, B. Lemming, P. Saini, Dr. P. A. Marshall,

Dr. P. W. Jurutka, Dr. C. E. Wagner

School of Mathematical and Natural Sciences

New College of Interdisciplinary Arts&Sciences

Arizona State University, 4701 W. Thunderbird Rd., Glendale, AZ 85306

(USA)

E-mail: Carl.Wagner@asu.edu

[b] Dr. T. L. Groy

Department of Chemistry and Biochemistry

College of Liberal Arts&Sciences

Arizona State University, 551 E. University Dr., Tempe, AZ 85287 (USA)

[c] G. M. Gray, J. C. Baso, Dr. A. van der Vaart

Department of Chemistry, University of South Florida

4202 E. Fowler Ave., CHE 205, Tampa, FL 33620 (USA)

Supporting information for this article is available on the WWW under

http://dx.doi.org/10.1002/cmdc.201200319.

ChemMedChem 0000, 00, 1 – 17

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

1

&

ÞÞ

These are not the final page numbers!

C. E. Wagner et al.

MED

RXR can be either liganded or unliganded. For exam-

ple, in the case of the RXR–VDR heterodimer, RXR is

believed to be unliganded.^[5]There are a few nuclear

receptors, such as the liver X receptor (LXR), that

form heterodimers with liganded RXR.^[6]Thus, RXR

often plays the role of the “master” partner, control-

ling the operation of several nuclear receptors

through liganded and unliganded heterodimers that effect

specific physiological equilibria via control of gene expres-

sion.^[7]

Additionally, compound 1 and several of its analogues have

been considered in non-insulin-dependent diabetes mellitus

(NIDDM) mouse models.^[16]Despite the ability of compound

1 to specifically activate RXR, versus RAR, the primary draw-

backs to treatment with 1 include hyperlipidemia, hypothyr-

oidism,^[17]and cutaneous toxicity. Many of these side effects

occur either because of antagonism of a non-permissive recep-

tor, as in the case of TR for hypothyroidism,^[18]or agonism of

a permissive receptor, as in the cases of LXR for hyperlipide-

mia^[19]and RAR for cutaneous toxicity,^[20]at the typical dose

concentration. Hence, there is ample motivation to explore

novel RXR agonists that may attenuate or avoid these side ef-

fects.

Notably, ligand-promoted RXR homodimer transcriptional

activity is suppressed for most cases where RXR is coordinated

in a heterodimer with an endogenous ligand-bound receptor

such as TR or VDR, and the TR and VDR partners in these RXR

heterodimers are termed “nonpermissive” partners for RXR.^[5]In

contrast to TR and VDR, RAR forms a heterodimer with RXR

when RAR binds to all-trans-retinoic acid, but the RXR partner

is still capable of binding 9-cis-RA, and both retinoids act in

synergy to promote RAR/RARE-mediated gene transcription.

When synthetic high-affinity RXR binding ligands (rexinoids) or

9-cis-RA are available, even in the presence of the TR and VDR

agonists for the TR-RXR or VDR-RXR heterodimers, the rexi-

noids divert RXR proteins from heterodimer formation, instead

promoting RXR homodimer formation and decreasing thyroid

hormone and 1,25(OH)₂D₃responsiveness. Altering the binding

ligand structure for a given nuclear receptor (NR), especially

the RXR master heteropartner ligand, produces specific NR

modulators (SNuRMs) that have unique properties that exert

novel influences on the NR activity.^[8]

There are a plethora of demonstrated RXR agonists modeled

on 1. For example, cyclopropyl dienoic acid 3,^[21]and several

novel aza-retinoids, of which compound 4^[22]is representative,

in addition to amide retinoids^[23]have all been disclosed. The

thiocarbamate bexarotene analogue 5^[24]induces apoptosis

when administered to leukemia HL-60 cells. Substituting pyri-

dine for one of the aromatic rings has led to several analogues

of 1, such as compound 6,^[25]and analogues of 1 possessing

unsaturation in the aliphatic ring, as in compound 7,^[26]have

also been reported. Studies describing the development of se-

lective RXR agonists containing aryl-trienoic acid moieties,

either alone,^[27]or locked by one^[28]or two^[29]ring systems,

were disclosed by Boehm et al., for which compound 8 is rep-

resentative of the latter. Acetal 9^[30]and compound 10^[7]are

both selective, potent RXR agonists, with the latter serving as

a model to design a potent RXR antagonist. Also, our group re-

cently synthesized an analogue of 1 bearing a fluorine atom

ortho to the carboxylic acid moiety, 2-fluoro-4-(1-(1,2,3,4-tetra-

hydro-1,1,4,4,6-pentamethylnaphthalen-7-yl)vinyl)benzoic acid

(11),^[31]that demonstrated slightly higher RXR activation in

Caco-2 colon cancer cells and possessed a slightly lower K_d

RXR-selective molecule (rexinoid) SNuRMs have been recent

targets for medicinal chemistry, as selective RXR activation

versus RAR appears to confer chemotherapeutic effects^[8]in

a number of cancers without inciting concomitant negative

side effects from RAR interaction.^[9]After extensive synthesis^[10]

of molecules modeled in part on the endogenous 9-cis-RA

(shown below), researchers at Ligand Pharmaceuticals Inc.,

demonstrated a highly selective RXR agonist, 4-[1-(3,5,5,8,8-

pentamethyltetralin-2-yl)ethenyl]benzoic acid (1),^[11]commonly

named bexarotene. A disilabexarotene compound 2,^[12]mod-

eled on 1 but substituting two silicon atoms for two carbon

atoms in the aliphatic ring system, demonstrated similar RXR

specific agonism.

Compound 1 has been FDA

approved to treat cutaneous T-

cell lymphoma (CTCL), has re-

cently been employed off-label

to treat lung cancer,^[13]and has

been analyzed as a treatment for

breast cancer,^[14]colon cancer,^[15]

and other uncontrolled cell pro-

liferation diseases, because pro-

moted expression of RXR regu-

lated genes appears to slow or

arrest cell proliferation as well as

predisposing cancerous cells to

apoptosis in the presence of

a

chemotherapeutic agent.^[14]

&2

&

www.chemmedchem.org

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemMedChem 0000, 00, 1 – 17

ÝÝ

These are not the final page numbers!

RXR-Selective Agonists

MED

than 1. Thus, our recent studies have systematically examined

by modeling, synthesis, and in vitro biological evaluation eight

novel analogues of 11 bearing different or additional halogen

atoms, exemplified by compounds 12–19.

Herein we report the synthesis of compounds 12–19 and

evaluate their potency and activity in biologically relevant sys-

tems including mammalian 2-hybrid (M2H) assays, RAR- and

RXR-response element (RARE and RXRE) transcriptional activa-

tion assays in cultured human cell lines, as well as in mutage-

nicity and apoptosis assays. We assessed these novel ana-

logues in comparison with compound 1 as well as its ketone

analogue 20.

tended to vacate the hydrophobic pocket, disrupting interac-

tions with the hydrophobic ring system of the ligand. This dis-

ruption led to a rotation of the ring system away from Ile268,

weakening the hydrophobic interactions with this residue. The

hydrophobic ring system was accommodated in a large hydro-

phobic pocket, with interactions between the ligand and

Ile345, Ile268, and Leu436.

Given the highly favorable docking energies of the class of

ortho-halogenated compounds, and the relative synthetic ac-

cessibility of this class, we decided to synthesize and test com-

pounds 12–18, as well as compounds 19–20. Ortho-halogenat-

ed compounds 12–17 were found to be within the best 20

binders for the library, and the

binding of compound 18 was

also predicted to be more favor-

able than bexarotene. The calcu-

lated relative binding free

energy for compounds 12–20 as

predicted by the docking studies

are shown in Figure 1. These cal-

culated binding free energies

were merely used to score the li-

gands; reported values are rela-

tive to the calculated binding

free energy of docked bexaro-

tene, with negative numbers in-

dicating better binding. An ex-

Results and Discussion

ample of a low-energy docking pose for compound 18 is

shown in Figure 2.

Molecular modeling

Computational docking studies were performed to help guide

the selection of ligands for synthesis. Analysis of the top bind-

ers in a large library of bexarotene-like analogues showed

a high preference for electron withdrawing groups ortho to

the carboxylic acid of the phenyl group; nearly all the top

binders in the docking studies contain a halogen or a nitro

group in this position. Structural comparisons of the docked

poses showed that the torsional angle between the phenyl

ring and the bridge head changed by ~15 to 308 in the ortho-

halogenated compounds relative to the non-halogenated com-

pounds. The effect of this rotation was a strengthening of the

interactions between Ile268 and the hydrophobic portion of

the ligand, as measured by decreased distances in the poses.

Halogenation did not change hydrogen bonding with Arg316,

however. The degree of rotation was inversely related to the

size of the halogen atom, with fluorine incurring the largest ro-

tation and stabilization of the hydrophobic group, and iodine

Chemistry

The synthesis of compounds 12–19 started with the radical di-

bromination of commercially available 3-chloro-4-methylben-

having

a smaller rotation and stabilization. Addition of

a second halogen amplified the effect. Replacement of the

phenyl group by a heteroatomic ring generally decreased

binding by disrupting interactions between the ring and

Leu309, Ala271, and Leu326. In the docking studies ligands

with methylene and cyclopropane bridge heads were predict-

ed to bind more favorably than ligands with ketone bridge

heads. Structural analyses of the docked poses showed that

methylene and cyclopropane bridge heads were bound in

a small hydrophobic pocket, while the ketone bridge head

Figure 1. Docking results for compounds 12–20; calculated binding free en-

ergies are relative to the calculated binding free energy of docked bexaro-

tene with negative numbers indicating better binding. Triangles represent

docks to 1MVC, circles to 1H9U; open symbols represent AutoDockTool

charges, and closed symbols OpenBabel charges. Compounds are listed

from left to right by increasing docking free energies averaged over the vari-

ous protocols.

ChemMedChem 0000, 00, 1 – 17

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemmedchem.org

&

3

&

ÞÞ

These are not the final page numbers!

C. E. Wagner et al.

MED

and 36 were converted smoothly to 38 and 39 at room tem-

perature and atmospheric pressure hydrogen, benzyl ester 37

required a higher temperature of 708C and higher pressure of

hydrogen (10–15 bar) that was safely achieved^[36]to give quan-

titative yield of 40. The carboxylic acids 38-40 were then quan-

titatively converted into the acid chlorides 41–43 with thionyl

chloride (Scheme 1).^[32]

A slightly different route was taken to prepare the iodinated

analogues 16 and 17. 4-(Methoxycarbonyl)-3-aminobenzoic

acid (44) was treated with sodium nitrite and hydrochloric acid

followed by potassium iodide to give 4-(methoxycarbonyl)-3-

iodobenzoic acid (45)^[37]in 77% yield, and acid 45 was convert-

ed into acid chloride 46 with thionyl chloride (Scheme 2).

Finally, 2,5-dimethyl-2,5-hexanediol was converted into 2,5-

dichloro-2,5-dimethylhexane (47),^[11,38]in 73% yield by treat-

ment with concentrated hydrochloric acid, and the dichloride

47 was reacted with toluene and catalytic aluminum chloride

to give 1,2,3,4-tetrahydro-1,1,4,4,6-pentamethylnaphthalene

(48)^[11,31]in 94% yield.

With compound 48 in hand, acid chlorides 41–43 and 46

were converted into ketones 49–52 according to the general

method of Boehm and co-workers.^[11]Iodo-ketone methyl ester

51 was observed to possess a small percentage (~7%) of the

chloro-ketone methyl-ester 49 by ¹H NMR analysis, and the

iodine atom may be labile under reflux conditions with alumi-

num trichloride. Compounds 49–52 were either saponified

with potassium hydroxide in methanol, followed by acidic

Figure 2. Low-energy docking pose for compound 18.

zoic acid (21) and 3-bromo-4-methylbenzoic acid (22) to give

dibromides 23 and 24, respectively, which were smoothly con-

verted into aldehydes 25 and 26 upon treatment with silver ni-

trate in ethanol and water according to the method reported

by Kishida and colleagues.^[32]We initially envisioned the syn-

thesis of the 3,5-difluoroformyl-

benzoic acid (27)^[33]to follow the

same route as used for alde-

hydes 25 and 26, beginning

with the known 3,5-difluoro-4-

methylbenzoic acid.^[34]However,

we chose to implement a three-

step, one-pot Bouveault alde-

hyde synthesis^[35]in which 3,5-di-

fluorobenzoic acid (28) was

treated with tert-butyl lithium,

followed by addition of dime-

thylformamide and hydrolysis

with hydrochloric acid to give 27

in 36% yield according to the

method of Anderson and co-

workers, instead.^[33]Continuing

with the general method of Kish-

Scheme 1. Reagents and conditions: a) NBS, cat. BPO, CCl₄, 858C (reflux), 36 h, 100%; b) AgNO₃, EtOH, H₂O, 558C,

45 min, 100%; c) 1. tBuLi, À788C, 30 min; 2. DMF, À788C, 1 h, then RT, 16 h; 3. HCl, 36% over three steps; d) NaH,

BnBr, DMF, RT, 5 h, 93–99%; e) NaClO₂, sulfamic acid, RT, 1 h, 82–90%; f) 1. SOCl₂, 848C, 1 h; 2. MeOH, Et₃N, RT, 1 h,

ida et al.,^[32]the carboxylic acid

aldehydes 25–27 were benzyl-

protected by treatment with 78–87% over two steps; g) H₂, Pd/C, RT, 24–48 h, 85–96%; h) SOCl₂, 858C, 1 h, 100%.

sodium hydride and benzyl bro-

mide to give benzyl ester alde-

hydes 29–31, respectively. The ester aldehydes 29–31 were

oxidized with sodium chlorite to give the benzyl ester acids

32–34, respectively, whose carboxylic acid moieties were con-

verted into methyl esters 35–37 by treatment with thionyl

chloride followed by methanol.^[32]The benzyl esters of 35–37

were converted into carboxylic acids 38-40 by treatment with

10% palladium on carbon and hydrogen, however, while 35

Scheme 2. Reagents and conditions: a) 1. HCl, NaNO₂, RT, 30 min; 2. KI, RT,

1 h, 40% over two steps; b) SOCl₂, 858C, 1 h, 100%.

&4

&

www.chemmedchem.org

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemMedChem 0000, 00, 1 – 17

ÝÝ

These are not the final page numbers!

RXR-Selective Agonists

MED

work-up, to give analogues 13, 15, 17, and 19 respectively, or

they were treated with triphenylphosphine methylide to give

alkenes 53–56, respectively. The bromo-ketone 15 also pos-

sessed minor amounts of unidentifiable impurities that could

not be removed by chromatography or crystallization. Alkenes

53–56 were then saponified with potassium hydroxide in

methanol, followed by acidic work-up, to give analogues 12,

14, 16, and 18, respectively (Scheme 3). While the iodo-ketone

Figure 3. a) The X-ray crystal structures of bexarotene analogues 12, 14, and

16 shown without hydrogen atoms, for clarity, with thermal ellipsoids at the

50% probability level. b) The X-ray crystal structures of bexarotene ana-

logues 52 and 18 shown without hydrogen atoms, for clarity, with thermal

ellipsoids at the 30% probability level. Compound 18 displayed twist-iso-

merism in the aliphatic ring. Hence, one twist isomer has been displayed.

with about 80, 60, and 20%, respectively, of the parent com-

pound’s efficiency independent of the cell line. They were also

found to be significantly better than the ethanol control vehi-

cle (P<0.05 for all, using one-tailed heteroscedastic t test).

Compound 18, however, was found to induce receptor binding

and homodimerization better than that of the parent com-

pound 1 to a degree considered statistically significant. More-

over, relative to compound 1, 18 was ~1.5-fold more potent in

its ability to induce RXR homodimerization than 1 in the Caco-

2 cells and ~3.5-fold better in HCT-116 cells (P<0.05 for all,

using one-tailed heteroscedastic t test). These results further

support that compounds modeled after 1 can be successfully

synthesized to possess RXR binding ability^[31]and agonistic

properties.

Scheme 3. Reagents and conditions: a) 48, AlCl₃, 558C (reflux), 15 min, 58–

89%; b) 1. KOH, MeOH, 858C (reflux), 1 h, 40–81%; 2. HCl; c) Ph₃PCH₃Br,

nBuLi, THF, RT, 1 h, 32–77%; d) 1. KOH, MeOH, 858C (reflux), 1 h, 34–91%;

2. HCl.

acid 17 contained ~9% of the chloro-ketone acid 13 and

could not be purified by column chromatography or attempt-

ed crystallization, the minor amount of chloro-alkene acid 12

could be removed from iodo-alkene 16 by crystallization from

ethyl acetate.

Novel analogues of 1 are able to direct RXR/RXRE

homodimer-mediated transcription

The X-ray crystal structures of compounds 12, 14, and 16

are shown in Figure 3a and the X-ray crystal structures of com-

pounds 18 and 52 are shown in Figure 3b.

The mammalian two-hybrid assay is useful as an initial screen

for RXR agonist induced homodimerization because of its ac-

cessibility, speed and sensitivity. However, because this assay

employs a system that uses a synthetic binding domain (BD)

and synthetic activating domain (AD), it is possible that within

this artificial context the RXR–agonist complex may have an al-

tered ligand or transcriptional co-activator affinity. Thus, it is

important to test our collection of potential RXR agonists for

the ability to mediate transcription in a more natural setting.

Hence, a second screening protocol included the transfection

of both Caco-2 and HCT-116 cells with human RXRa and an au-

thentic RXRE from the naturally occurring responsive element

in the rat cellular retinol binding protein II gene^[39]linked to

a luciferase reporter gene. The results in Figure 5 demonstrate

that compounds 12, 14, 18, and 19 are able to activate RXR

homodimer-mediated transcription significantly better than

that of the ethanol vehicle (for compounds 12, 14, and 18 in

Caco-2 cells, using a one-tailed heteroscedastic t test: P<0.05;

Biological assays and rationale

A mammalian two-hybrid assay demonstrates that several

novel analogues bind RXR in an agonist fashion

Biological evaluation of the synthetic analogues described

above (compounds 12–19) was first carried out in a mammalian

two-hybrid assay in human colon cancer (Caco-2) cells then re-

peated in a different colorectal carcinoma line (HCT-116) (Fig-

ure 4a and b, respectively). This assay tests the ability of the

analogue to bind to the recombinant human RXR receptor and

induce homodimerization as measured by luciferase output.

Four compounds were identified in this initial evaluation

whose agonist activity ranged from 20 to 400% of the binding

of compound 1 in the two separate cell lines. More specifically

12, 14, and 16 are able to bind and mediate homodimerization

ChemMedChem 0000, 00, 1 – 17

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemmedchem.org

&

5

&

ÞÞ

These are not the final page numbers!

C. E. Wagner et al.

MED

sess EC₅₀values that mirror the results of the mam-

malian two-hybrid analysis in Figure 4, and the RXRE-

based assays in Figure 5.

Analysis of RAR agonist activity

Because compound 1 is known to have some “residu-

al” retinoic acid receptor (RAR) agonist activity^[31]we

evaluated this group of analogues for activation of

this closely related nuclear receptor. This assay uses

the expression of human RAR and a retinoic acid re-

sponsive element (RARE)-luciferase reporter gene,

and it demonstrates (Table 2) that compounds 12,

14, and 18 are similar to 1 (P>0.05 for all, using

a

one-tailed heteroscedastic t test). Table 2 also

shows that 16 and 19 possess significantly lower RAR

binding affinity (P<0.05 for both, using a one-tailed

heteroscedastic t test). all-trans-retinoic acid was used

as the positive control in this experiment because it

is the endogenous ligand for RAR. Taken together,

these results not only further support our previous

findings^[31]that variation of 1 with a halogen atom

on the aromatic that bears the carboxylic acid may

decrease the activation of RAR (analogues 16 and 19)

or increase its capacity to bind and activate RXR

(compound 18), but they also strongly suggest a spe-

cific periodic trend for the substitution of a proton

that may relate to the analogue’s degree of agonist

activity.

Figure 4. Evaluation of potential RXR-selective agonists via a mammalian two-hybrid

assay in two different types of human colon cancer cells, a) Caco-2 and b) HCT-116. Both

cell lines were transfected with pCMV-BD-hRXR binding domain vector (BD), pCMV-AD-

hRXR activation domain (AD), pFR-Luc reporter gene containing BD binding sites, and

a renilla control plasmid. Cells were transfected for 7 h using a liposome-mediated trans-

fection protocol then exposed to either the ethanol vehicle or 10^À7m compound 1 or

the indicated analogue. After 24 h the cells were lysed and a luciferase assay was com-

pleted. Analogue-dependent RXR binding and homodimerization, as measured by lucifer-

ase output, was compared with the parent compound 1 (value set to 1.0).

Apoptosis in a CTCL system

As previously reported,^[31]bexarotene and several of

its analogue are capable of inducing apoptosis. We

for these compounds in HCT-116 cells using a one-tailed heter-

were interested in determining the pro-apoptotic ability of

these new analogues in comparison with compound 1. As

compound 1 has been effectively employed in the treatment

of CTCL because it induces apoptosis in the T-lymphocyte, we

assayed HuT-78 lymphocytes treated with compound 1 or ana-

logues for caspase 3 and 7 activity, a hallmark of apoptosis,

and compared caspase activity to cells treated with ethanol ve-

hicle and sodium butyrate (apoptotic positive control)

oscedastic t test: P<0.01; and for compound 19 in both cell

lines, using a one-tailed heteroscedastic t test: P<0.01).

In the Caco-2 cells, 16 was a weak agonist, while 17 was not

able to activate transcription when compared with ethanol,

but both analogues did display significant activity in the HCT-

116 colorectal carcinoma (using a one-tailed heteroscedastic

t test: P<0.01 for 16 and P<0.001 for 17).

Importantly, compounds 12, 14, 16 and 18 displayed activity

in both the mammalian two-hybrid assay (Figure 4) and the

RXRE assay (Figure 5), thus revealing that both assays are not

only valid, but generate consistent and complementary data

when evaluating RXR agonists.

Table 1. Determination of EC₅₀values.^[a]

Compd

EC₅₀[nm]

Bexarotene 1

55Æ6

90Æ14

150Æ18

280Æ34

34Æ6

12

14

16

18

19

Determination of ligand–receptor binding affinity

To determine the relative binding affinity and effectiveness of

our most active analogues, a mammalian two-hybrid assay was

performed with ligand concentrations ranging from 10ꢁ

10^À10m up to 0.5ꢁ10^À5m and EC50 values were calculated

(Table 1). Notably, the difluorobexarotene analogue (18) has an

EC₅₀value of 34Æ6 nm in HCT-116 cells versus an EC₅₀value of

55Æ6 nm for bexarotene (1). All of the other analogues pos-

550Æ67

[a] EC₅₀values were determined from full dose–response curves ranging

from 10^À10to 10^À5m in transfected HCT-116 cells using an RXR mammali-

an two-hybrid system as described in the Experimental Section. Values

represent the meanÆSD of n=2 independent experiments with triplicate

samples in each.

&6

&

www.chemmedchem.org

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemMedChem 0000, 00, 1 – 17

ÝÝ

These are not the final page numbers!

RXR-Selective Agonists

MED

Figure 6. Apoptosis analysis in a CTCL cell line. HuT-78 cells were treated for

48 h with analogue, ethanol vehicle, or sodium butyrate as a positive con-

trol. Cells were analyzed for apoptosis via a caspase assay (Promega Cas-

pase-Glo 3/7 assay) according to manufacturer’s instructions. Data is an aver-

age of two independent experiments with triplicate samples in each and

plotted as a percentage of sodium butyrate activity (set at 100%).

Conclusions

Figure 5. Detection of potential RXR agonists via an RXRE-luciferase based

system using human colon cancer cells, a) Caco-2 and b) HCT-116. Both cell

lines were transfected with hRXRa, an RXRE luciferase reporter gene, renilla

control plasmid, and carrier DNA (pTZ18U). Cells were transfected for 7 h

using a liposome-mediated transfection protocol then exposed to either the

ethanol vehicle or 10^À7m compound 1 or the indicated analogue. After 24 h

the cells were lysed and a luciferase assay was completed. Analogue-depen-

dent, RXR-mediated transcription, as measured by luciferase output, was

compared with the parent compound 1 (value set to 1.0).

Here we report the modeling, synthesis, and biological evalua-

tion of several analogues of compound 1, in an extension of

our earlier work.^[31]We have shown that the addition of two

fluorine atoms ortho to the carboxylic acid of 1 increases the

ability of analogue 18 to activate RXR. We have identified sev-

eral new halogenated analogues of 1 that have apparent bind-

ing and biological activity similar to the parent compound,

and even follow a periodic trend, with one (compound 18)

that capitalizes on the finding that adding one fluorine ortho

to the carboxylic acid of 1 increases binding and activation of

RXR.^[31]With the fact that several RXR selective agonists are

being explored to treat many conditions through biological

pathways impacted by RXR, now including the demonstrated

up-regulation of the apoE gene and the facilitated clearance of

plaques by 1 in mouse models of Alzheimer’s disease,^[40]there

is motivation to develop new RXR agonists. Our results indi-

cate that novel analogues of 1 that substitute halogen atoms

on the aromatic ring bearing the carboxylic acid can likely

serve as effective, and perhaps more potent, ligands for RXR,

which may also have decreased RAR agonist activity (Table 2);

thus, these compounds may also possess less detrimental side

effects in cutaneous T-cell lymphoma patients.

Table 2. Quantitation of RAR agonist activity.^[a]

Compd

Agonist activity [%]

(100 nm)

(1 mm)

Bexarotene 1

21Æ4

13Æ2

5Æ1

8Æ1

4Æ1

19Æ3

14Æ2

5Æ1

9Æ2

8Æ2

12

14

16

18

19

[a] RAR agonist activity was derived from an RAR/RARE reporter system in

transfected HEK-293 cells treated with test compound or all-trans-retinoic

acid (RA) at either 100 nm or 1 mm. RAR agonist activity is defined as the

activity with test compound (or reference 1) divided by the activity with

all-trans-RA expressed as a percentage. Values represent the meanÆSD

of n=3 independent experiments with triplicate samples in each.

Experimental Section

(Figure 6). Similar to previously published results,^[31]bexarotene

(1) and all of the analogues were better than vehicle control at

inducing apoptosis (P<0.01 using one-tailed heteroscedastic

t test), after a 48 hour treatment period in our CTCL cells.

Docking Studies

Docking studies using AutoDock 4.2^[41]were performed using the

X-ray structures of human RXRa in complex with BMS 649^[42]and

RXRb in complex with LG100268,^[43]PDB access codes 1MVC and

1H9U, respectively. In both cases, the Arg316 and Ile268 protein

residues (numbering as in 1MVC) were treated flexible. Arg316 was

selected to enable hydrogen bonding with carboxylate groups on

the phenyl moiety, an important interaction identified in earlier

ChemMedChem 0000, 00, 1 – 17

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemmedchem.org

&

7

&

ÞÞ

These are not the final page numbers!

C. E. Wagner et al.

MED

studies.^[11,31]Ile268 was selected, as structural overlays of 33 differ-

ent ligand-bound RXR structures in the PDB showed large motions

of Ile268, and in all cases Ile268 made significant interactions with

hydrophobic portions of the ligands. Atomic charges generated by

AutoDockTools^[41]were frequently overpolarized for nitro groups

(leading to a net negative charge on the nitro group), while charg-

es generated by OpenBabel 2.3.0^[44]did not have this artifact.

Therefore, all docks were performed with both charge models, and

results for nitro-compounds were omitted for the AutoDockTools-

generated charges. Docking was performed with the Lamarckian

genetic algorithm using a maximum of 25ꢁ10⁶energy evaluations

per docking. The number of docks was set to 250 for three tor-

sions; this number was increased by 50 for each additional torsion,

up to a maximum of 400 for six or more torsions. The docking was

performed with a large library of bexarotene-like analogues. These

analogues differed in substituents on the phenyl ring (including

halogenated, carboxylated, hydroxylated and nitrated rings) and

fused phenyl ring (including methylated, aminated, nitrated, and

halogenated rings), as well as changes in the identity of the bridge

head between the ring systems (including methylene, ketone, and

cyclopropane bridge heads, as well as the absence of a bridge

head), and the identity of the aliphatic part of the fused ring

system. Calculated AutoDock binding free energies were used to

score the ligands.

RAR/RARE-agonist activity assay: An embryonic kidney cell line,

HEK-293, was used for the RARE transcription assay. The cells were

plated at a concentration of 7ꢁ10⁴cells per well and maintain as

described above. The cells were allowed to incubate over night to

ensure attachment to the plate surface. The transfection protocol

called for 20 ng of renilla null control plasmid to monitor transfec-

tion efficiency, 30 ng of pTZ18U carrier DNA plasmid, 50 ng of the

pCMX-human RARa expression vector, and 250 ng of pTK-DR5(X2)-

Luc plasmid. Specifics pertaining to this RARE-containing reporter

vector have been described previously as well as the sequence of

the double RARE.^[45]The cells were incubated for 7 h and then

treated with ethanol, all-trans-retinoic acid, or analogues at final

concentration ranging from 10^À6m to 10^À7m for 24 h. After the in-

cubation period cell were lysed and the same luciferase assay was

completed that was previously described.

Apoptosis assay: Apoptotic activity was assessed by the Caspase-

Glo 3/7 Assay (Promega, Madison, WI) according to the manufac-

turer’s instructions. The Caspase-Glo 3/7 Assay is based on the

cleavage of the DEVD sequence of a luminogenic substrate by cas-

pases 3 and 7 which results in a luminescent signal. HuT-78

(human T-cell lymphoma) cells were distributed (1ꢁ10⁴cells per

well) in white-walled 96-well microplates (Corning, NY) in 100 mL of

medium and incubated with 500 mm sodium butyrate (NaBu),

10 mm bexarotene (Bex) or rexinoid analogues for 48 h. The Cas-

pase-Glo 3/7 Reagent was then added to each well and incubated

for an additional 1 h at RT. The luminescence was measured in a lu-

minometer (Safire2, Tecan, US). NaBu, a known inducer of apopto-

sis in HuT-78, was used as a positive control. Each treatment group

was dosed in triplicate, and at least two independent experiments

were performed with triplicate samples in each. Numbers were

standardized to sodium butyrate (set at 100%).

Biological evaluation

Mammalian two-hybrid assay: Caco-2 colorectal carcinoma cells

were plated overnight at 7ꢁ10⁴cells per well in a 24 well plate

and kept in minimum essential media (MEM) (Invitrogen, Carlsbad

CA) enhanced with 20% fetal bovine serum (FBS) (Invitrogen),

1 mm sodium pyruvate (Invitrogen), 100 mgmL^À1streptomycin, and

100 UmL^À1penicillin. HCT-116 colorectal carcinoma cells were

plated overnight at 7ꢁ10⁴cells per well in a 24 well plate and

kept in DMEM enhanced with 10% FBS (Invitrogen), 1 mm sodium

pyruvate (Invitrogen), 100 mgmL^À1streptomycin, and 100 UmL^À1

penicillin. Both cell lines were co-transfected using a human RXR

binding domain vector, a hRXR activation domain (AD), a luciferase

reporter gene containing BD binding sites, and a renilla control

plasmid. A liposome-mediated transfection was completed accord-

ing to the manufacturer’s protocol using 2 mL per well of Express-

In transfection reagent (Thermo Fisher Scientific, Lafayette, CO)

and allowed to incubate for 7 h. The cells were then treated with

Mutagenicity: We tested the compounds for mutagenicity as in

Wagner et al.^[31]None of the compounds are mutagenic. This assay

used a yeast strain, D7, that is genetically engineered to change

phenotype upon a genotype change.^[46]We used this strain to test

the mutagenicity of the compounds by solubilizing the com-

pounds in DMSO and performing a dose–response curve with the

highest concentration being 0.15% w/v comparing with DMSO

control for mutagenicity, scored as a phenotype change on agar

plates.^[47]

Chemistry

ethanol vehicle, 1, or analogues at a final concentration of 10^À7

m

and incubated for 24 h. The amount of rexinoid activity was mea-

sure by luciferase output using a dual-luciferase reporter assay

system according to the manufacturer’s protocol (Promega, Madi-

son, WI) in a Sirus FB12 luminometer (Berthold Detection Systems,

Zylux Corporation, Huntsville, AL). Several independent assays

were conducted with triplicate samples for each treatment group.

Instrumentation: A 400 MHz Bruker spectrometer was used to ac-

1

quire H NMR and ¹³C NMR spectra. Chemical shifts (d) are listed in

ppm against residual non-deuterated solvent peaks in a given deu-

terated solvent (e.g., CHCl₃in CDCl₃) as an internal reference. Cou-

pling constants (J) are reported in Hz, and the abbreviations for

splitting include: s, single; d, doublet; t, triplet; q, quartet; p,

pentet; m, multiplet; br, broad. All ¹³C NMR spectra were acquired

on a Bruker instrument at 100.6 MHz. Chemical shifts (d) are listed

in ppm against deuterated solvent carbon peaks as an internal ref-

erence. High-resolution mass spectra were recorded using either

a JEOL GCmate(2004), a JEOL LCmate(2002) high-resolution mass

spectrometer or an ABI Mariner (1999) ESI-TOF mass spectrometer.

HPLC traces were obtained on an Agilent 1100 LC with a Phenom-

enex Kinetex C18 10 cm by 2.1 mm column with 2.6u solid core

particles.

RXRE-mediated transcription assay: The RXRE assays were complet-

ed using both Caco-2 and HCT-116 cells plated at 7ꢁ10⁴cells per

well in a 24 well plate and maintained as described above. The

cells were co-transfected using 250 ng of RXRE-luciferase reporter

gene (RXRE from the naturally occurring responsive element in the

rat cellular retinol binding protein II gene), 20 ng of the renilla con-

trol plasmid, 50 ng of human pSG5-RXRa, and 100 ng pTZ18U car-

rier DNA plasmid and 2 mL per well of Express-In was again used

for the liposome mediated delivery. The cells were incubated for

7 h post-transfection and then treated with ethanol, or 10^À7m of

either the parent compound or the indicated analogue. After a 24-

hour incubation period the amount of retinoid activity was mea-

sured using the same luciferase assay described above.

General procedures: Tetrahydrofuran, methylene chloride, diethyl

ether, and benzene were dried by filtration through alumina ac-

cording to the procedure described by Grubbs.^[48]Removal of vola-

tile solvents transpired under reduced pressure using a Bꢂchi

&8

&

www.chemmedchem.org

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemMedChem 0000, 00, 1 – 17

ÝÝ

These are not the final page numbers!

RXR-Selective Agonists

MED

rotary evaporator and is referred to as removing solvents in vacuo.

Thin-layer chromatography was conducted on pre-coated

(0.25 mm thickness) silica gel plates with 60 F₂₅₄indicator (Merck).

Column chromatography was conducted using 230–400 mesh

silica gel (E. Merck reagent silica gel 60). All tested compounds

were analyzed for purity by combustion analysis through Columbia

Analytical Services (formerly Desert Analytics in Tucson, AZ, USA)

and were found to be >95% pure.

164.4, 135.2, 134.8, 134.7, 132.4, 132.3, 132.2, 128.7, 128.6, 127.6,

67.6 ppm; IR (neat): n˜ =2964, 1724, 1685, 1603, 1557, 1484,

1455 cm^À1

;

LC–FAB-MS: m/z [M⁺+H] calcd for C₁₅H₁₂O₄Cl:

291.0424, found: 291.0434.

4-Benzyl 1-methyl 2-chlorobenzene-1,4-dioate (35): Compound

35 was synthesized according to the method of Kishida and co-

workers.^[32]To a 100 mL round-bottom flask charged with com-

pound 32 (5.5 g, 18.9 mmol) was added SOCl₂(16.0 mL, 220 mmol)

and the reaction solution was held at reflux for 1 h in an oil bath

at 848C. The reaction solution was cooled to RT and the excess

SOCl₂was removed in vacuo to give crude 4-chlorocarbonyl-2-

chlorobenzoic acid benzyl ester. The crude 4-chlorocarbonyl-2-

chlorobenzoic acid in dry toluene (8 mL) was added dropwise to

a solution of Et₃N (5.2 mL, 37 mmol) in MeOH (53 mL, 1.3 mol) over

10 min. The reaction solution was stirred 1 h and poured into 1n

HCl (150 mL) and extracted with EtOAc (80 mL, thrice). The com-

bined organic extracts were washed with brine, dried over Na₂SO₄,

and removed in vacuo to give crude 35. Crude 35 was purified by

column chromatography (150 mL SiO₂, hexanes/EtOAc 95:5) to

Benzyl-3-chloro-4-formylbenzoate (29): Compound 29 was syn-

thesized according to the methods of Kishida and co-workers.^[32]To

a 500 mL round-bottom flask charged with 3-chloro-4-methylben-

zoic acid (21) (3.24 g, 19.0 mmol) was added NBS (8.00 g,

44.9 mmol), benzoyl peroxide (0.23 g, 0.95 mmol), and CCl₄

(37 mL). The reaction solution was heated at reflux under magnetic

stirring for 36 h, cooled to RT, and solids were filtered and washed

with CCl₄(~20 mL). The filtrate solvent was removed in vacuo and

the crude 4-(dibromomethyl)-3-chlorobenzoic acid (23) was dried

on high vacuum and used without further purification. To

a 500 mL round-bottom flask charged with crude 23 (6.23 g,

19.0 mmol) was added EtOH (48 mL), and a solution of AgNO₃

(6.64 g, 39.1 mmol) in warm water (9 mL) was added dropwise

while the reaction solution was stirred in an oil bath preheated to

50–558C. Upon addition of the AgNO₃solution, a green precipitate

formed. After stirring at 508C for 45 min, the reaction solution was

cooled to RT and filtered to remove the green precipitate. The fil-

trate solvent was concentrated in vacuo, extracted with EtOAc, and

the combined organic extracts were washed with water and brine,

dried over Na₂SO₄and removed in vacuo to give crude 3-chloro-4-

formylbenzoic acid (25) (3.51 g, 100%) that was used without fur-

ther purification. To a 500 mL round-bottom flask charged with 25

(3.51 g, 19.0 mmol) was added dry dimethylformamide (37 mL) and

a 60 wt% suspension of NaH in mineral oil (0.93 g, 23 mmol) in

small aliquots over 20 min. The reaction solution was stirred an ad-

ditional 20 min, and benzyl bromide (2.8 mL, 23 mmol) was added

to the red heterogeneous solution. After stirring 5 h, the reaction

solution had become homogeneous, and it was poured into 1n

HCl (100 mL), extracted with EtOAc (2ꢁ100 mL), and the combined

organic extracts were washed with saturated NaHCO₃(75 mL) and

brine, dried over Na₂SO₄, and removed in vacuo to give crude 29.

Crude 29 was purified by column chromatography (150 mL SiO₂,

hexanes/EtOAc 4:1) to give 29 (5.2 g, 99%) as a colorless oil:

¹H NMR (400 MHz, CDCl₃): d=10.52 (s, 1H), 8.15 (dd, J=8.0, 1.6,

1H), 7.97–8.04 (m, 2H), 7.37–7.48 (m, 5H), 5.39 ppm (s, 2H); IR

(neat): n˜ =2968, 2855, 2656, 2560, 1724, 1687, 1605, 1556,

1485 cm^À1; GC–MS: m/z [M⁺] calcd for C₁₅H₁₁O₃Cl: 274.0397, found:

274.0390.

1

give pure 35 as a white solid (5.05 g, 87%), mp: 46–488C: H NMR

(400 MHz, CDCl₃): d=8.12 (d, J=1.6, 1H), 7.98 (dd, J=8.0, 1.6, 1H),

7.85 (d, J=8.0, 1H), 7.36–7.46 (m, 5H), 5.38 (s, 2H), 3.95 ppm (s,

3H); ¹³C NMR (100.6 MHz, CDCl₃): d=165.5, 164.4, 135.3, 134.0,

133.8, 133.7, 132.0, 131.2, 128.7, 128.6, 128.4, 127.6, 67.4, 52.7 ppm;

IR (neat): n˜ =3033, 2958, 1735, 1716, 1560, 1498, 1482 cm^À1; GC–

MS: m/z [M⁺] calcd for C₁₆H₁₃O₄Cl: 304.0502, found: 304.0498.

4-(Methoxycarbonyl)-3-chlorobenzoic acid (38): Compound 38

was synthesized according to the methods of Kishida and co-work-

ers.^[32]A 3-neck 250 mL round-bottom flask charged with 35

(4.94 g, 16.2 mmol), 10% Pd/C (0.499 g), EtOH (28.0 mL), and EtOAc

(28.0 mL) was evacuated and back-filled with hydrogen gas from

a balloon three times, and the reaction solution was allowed to stir

under hydrogen at RT overnight. The reaction solution was filtered

through Celite, and the solvents were removed in vacuo to give

crude 38 (2.97 g, 85%) as a white crystalline solid that was used

without further purification. A small sample of crude 38 was puri-

fied by recrystallization from hot EtOAc to give pure 38, mp: 156–

1

1578C: H NMR (400 MHz, [D₆]DMSO): d=13.64 (brs, 1H), 7.99 (d,

J=1.6, 1H), 7.95 (dd, J=8.0, 1.6, 1H), 7.89 (d, J=8.0, 1H),

3.89 ppm (s, 3H); ¹³C NMR (100.6 MHz, [D₆]DMSO): d=165.4, 165.1,

134.9, 133.7, 131.8, 131.2, 130.9, 128.0, 52.8 ppm; IR (neat): n˜ =

2961, 2824, 2545, 1717, 1687, 1557, 1488 cm^À1; LC–APCI-MS: m/z

[M⁺] calcd for C₉H₇O₄Cl: 214.0033, found: 214.0027.

Methyl 2-chloro-4-(1-(3,5,5,8,8-pentamethyl-5,6,7,8-tetrahydro-

naphthalen-2-yl)carbonyl)benzoate (49): Compound 49 was syn-

thesized according to the method of Boehm and co-workers.^[11]

Methyl 4-(chlorocarbonyl)-2-chlorobenzoate (41) was synthesized

by refluxing 4-(methoxycarbonyl)-3-chlorobenzoic acid (38) (1.35 g,

6.29 mmol) in SOCl₂(10.0 mL, 137 mmol) in a 100 mL one-neck

round-bottom flask fitted with a water-cooled reflux condenser.

Excess SOCl₂was removed in vacuo to give crude 41 as an off-

white solid, and this solid was dissolved in dry benzene (~20 mL)

and evaporated to dryness three times to remove residual SOCl₂.

The acid chloride 41 was dried on high vacuum to remove residual

benzene. To a 2-neck, 50 mL round-bottom flask equipped with

a reflux condenser and magnetic stir-bar was added 48 (1.38 g,

6.82 mmol) followed by a solution of crude acid chloride 41

(6.29 mmol) in CH₂Cl₂(15 mL). AlCl₃(2.0 g, 15 mmol) was added to

the reaction solution at RT slowly, with stirring, and the reaction so-

lution turned from colorless to red accompanied by the evolution

of gas and heat. The reaction was stirred for 5 min then heated at

reflux for 15 min. The reaction was judged to be complete by TLC,

4-((Benzyloxy)carbonyl)-2-chlorobenzoic acid (32): Compound 32

was prepared according to the method of Kishida and co-work-

ers.^[32]To a 100 mL round-bottom flask charged with 29 (5.22 g,

19.0 mmol), sulfamic acid (1.86 g, 19.2 mmol), water (20 mL), and

ACN (15 mL) was added a solution of 80% NaClO₂(1.79 g,

19.8 mmol) in water (10 mL). After stirring for 1 h, the reaction so-

lution was poured into saturated Na₂SO₃(25 mL) and 1n HCl

(50 mL), and the resulting solution was extracted with EtOAc

(50 mL, thrice). The combined organic extracts were washed with

brine, dried over Na₂SO₄, and removed in vacuo to give crude 32

(4.97 g, 90%) that was used without further purification. A small

sample was purified by column chromatography (25 mL SiO₂, hex-

anes/EtOAc 1:1) to give pure 32 as a white powder, mp: 120–

1228C: ¹H NMR (400 MHz, CDCl₃): d=10.71 (brs, 1H), 8.17 (d, J=

1.6, 1H), 8.05 (d, J=8.0, 1H), 8.02 (dd, J=8.0, 1.6, 1H), 7.36–7.47

(m, 5H), 5.40 ppm (s, 2H); ¹³C NMR (100.6 MHz, CDCl₃): d=170.1,

ChemMedChem 0000, 00, 1 – 17

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

www.chemmedchem.org

&

9

&

ÞÞ

These are not the final page numbers!

C. E. Wagner et al.

MED

and the solution was poured into an ice solution (25 mL) acidified

with a 20% HCl solution (8 mL) and EtOAc was added (13 mL). The

aqueous and organic layers were separated, and the aqueous layer

was extracted with EtOAc (2ꢁ15 mL). The combined organics were

washed with water and brine, dried over Na₂SO₄, filtered and con-

centrated to give crude 49. Crude 49 was purified by column chro-

matography (250 mL SiO₂, hexanes/EtOAc 95:5 to 92.5:7.5) to give

¹³C NMR (100.6 MHz, CDCl₃): d=170.5, 147.8, 146.8, 144.7, 142.5,

137.1, 135.0, 132.6, 132.5, 129.3, 128.2, 128.0, 126.5, 124.9, 118.2,

35.1, 34.0, 33.9, 31.9, 31.8, 19.9 ppm; IR (neat): n˜ =2958, 1694,

1596, 1488 cm^À1; LC–APCI-MS: m/z [M⁺+H] calcd for C₂₄H₂₈O₂Cl:

383.1778, found: 383.1778; Anal. calcd for C₂₄H₂₇O₂Cl: C 75.28; H

7.11; Cl 9.26, found: C 75.20; H 6.93; Cl 9.60.

2-Chloro-4-(1-(3,5,5,8,8-pentamethyl-5,6,7,8-tetrahydronaphtha-

len-2-yl)carbonyl)benzoic acid (13): Compound 13 was synthe-

sized following the method of Boehm and co-workers.^[11]To

a 100 mL round-bottom flask charged with 49 (0.353 g, 0.88 mmol)

and MeOH (4 mL) was added a 5m aq KOH (0.4 mL, 2.0 mmol). A

reflux condenser was fitted to the round-bottom flask and the re-

action solution was held at reflux and monitored by TLC. After 1 h

at reflux, the reaction solution was cooled to RT and quenched

with 20% HCl (20 mL). The precipitate was filtered and washed

with water to give crude 13 (0.335 g, 98%). Crude 13 was purified

by column chromatography (25 mL SiO₂, hexanes/EtOAc 9:1) to

give 13 (0.13 g, 40%) as a white crystalline solid, mp: 174–1758C:

¹H NMR (400 MHz, CDCl₃): d=8.07 (d, J=8.0, 1H), 7.92 (d, J=1.6,

1H), 7.74 (dd, J=8.0, 1.6, 1H), 7.27 (s, 1H), 7.23 (s, 1H), 2.37 (s,

3H), 1.70 (s, 4H), 1.32 (s, 6H), 1.22 ppm (s, 6H); ¹³C NMR

(100.6 MHz, CDCl₃): d=196.0, 169.9, 149.0, 142.6, 142.1, 134.9,

134.7, 133.7, 132.7, 132.1, 131.6, 129.7, 128.7, 127.9, 34.8, 34.7, 34.4,

33.9, 31.6, 31.5, 20.1 ppm; IR (neat): n˜ =2955, 2926, 1704, 1667,

1543, 1458 cm^À1; LC–APCI-MS: m/z [M⁺+H] calcd for C₂₃H₂₆O₃Cl:

385.1571, found: 385.1564; Anal. calcd for C₂₃H₂₅O₃Cl: C 71.77; H

6.55; Cl 9.21, found: C 71.45; H 6.30; Cl 9.2.

1

49 (2.22 g, 88%) as a white, crystalline solid, mp: 97–988C: H NMR

(400 MHz, CDCl₃): d=7.89 (d, J=8.0, 1H), 7.88 (d, J=1.6, 1H), 7.70

(dd, J=8.0, 1.6, 1H), 7.25 (s, 1H), 7.21 (s, 1H), 3.96 (s, 3H), 2.35 (s,

3H), 1.69 (s, 4H), 1.31 (s, 6H), 1.20 ppm (s, 6H); ¹³C NMR

(100.6 MHz, CDCl₃): d=196.0, 165.7, 148.8, 142.0, 141.8, 134.8,

133.9, 133.7, 133.3, 132.3, 131.1, 129.6, 128.6, 127.8, 52.7, 34.8, 34.7,

34.3, 33.8, 31.6, 31.5, 20.0 ppm; IR (neat): n˜ =2961, 2864, 1738,

1665, 1547, 1484 cm^À1; GC–MS: m/z [M⁺] calcd for C₂₄H₂₇O₃Cl:

398.1649, found: 398.1657.

Methyl 2-chloro-4-(1-(1,2,3,4-tetrahydro-1,1,4,4,6-pentamethyl-

naphthalen-7-yl)vinyl)benzoate (53): Compound 53 was synthe-

sized according to the method of Boehm and co-workers.^[11]To

a 20-dram vial containing 49 (0.830 g, 2.08 mmol) and dry THF

(3 mL) at RT with a Teflon magnetic stir-bar was slowly added a tri-

phenylphosphonium methylide solution prepared as follows: to

a 100 mL round-bottom flask equipped with a Teflon magnetic stir-

bar and containing dry THF (2.0 mL) was added iPr₂NH (0.66 mL,

4.67 mmol) and a 2.5m solution of n-butyl lithium in hexanes

(1.7 mL, 4.25 mmol), and the solution was stirred for 30 min at RT

at which point, methyl triphenylphosphonium bromide (1.13 g,

3.19 mmol) was added and the solution was stirred an additional

20 min to provide a homogeneous dark yellow ylide solution. The

reaction was monitored by TLC, and when the reaction was

judged to be complete, the reaction solution was poured into

water (50 mL) and the aqueous solution was extracted with EtOAc

(2ꢁ50 mL). The combined organic extracts were washed with

water and brine, dried over Na₂SO₄, and concentrated in vacuo to

give crude 53 which was purified by column chromatography

(25 mL SiO₂, hexanes/EtOAc 97.5:2.5) to give 53 (0.283 g, 34%) as

Benzyl-3-bromo-4-formylbenzoate (30): Compound 30 was syn-

thesized according to the methods of Kishida and co-workers.^[32]To

a 500 mL round-bottom flask charged with 3-chloro-4-methylben-

zoic acid (22) (8.17 g, 38.0 mmol) was added NBS (16.10 g,

90.45 mmol), benzoyl peroxide (0.45 g, 1.8 mmol), and CCl₄

(74 mL). The reaction solution was heated at reflux under magnetic

stirring for 36 h, cooled to RT, and solids were filtered and washed

with CCl₄(~20 mL). The filtrate solvent was removed in vacuo and

the crude 4-(dibromomethyl)-3-bromobenzoic acid (24) was dried

on high vacuum and used without further purification. To

a 500 mL round-bottom flask charged with crude 24 (14.16 g,

38.0 mmol) was added EtOH (96 mL), and a solution of AgNO₃

(13.28 g, 78.18 mmol) in warm water (18 mL) was added dropwise

while the reaction solution was stirred in an oil bath preheated to

50–558C. Upon addition of the AgNO₃solution, a green precipitate

formed. After stirring at 508C for 45 min, the reaction solution was

cooled to RT and filtered to remove the green precipitate. The fil-

trate solvent was concentrated in vacuo, extracted with EtOAc, and

the combined organic extracts were washed with water and brine,

dried over Na₂SO₄and removed in vacuo to give crude 3-bromo-4-

formylbenzoic acid (26) (8.7 g, 100%) that was used without fur-

ther purification. To a 500 mL round-bottom flask charged with 26

(8.7 g, 38.0 mmol) was added dry dimethylformamide (74 mL) and

a 60 wt% suspension of NaH in mineral oil (1.86 g, 46.5 mmol) in

small aliquots over 20 min. The reaction solution was stirred an ad-

ditional 20 min, and benzyl bromide (5.60 mL, 46.8 mmol) was

added to the red heterogeneous solution. After stirring 5 h, the re-

action solution had become homogeneous, and it was poured into

1n HCl (200 mL), extracted with EtOAc (2ꢁ100 mL), and the com-

bined organic extracts were washed with saturated NaHCO₃

(75 mL) and brine, dried over Na₂SO₄, and removed in vacuo to

give crude 30. Crude 30 was purified by column chromatography

(150 mL SiO₂, hexanes/EtOAc 4:1) to give 30 (12.1 g, 99%) as a col-

1

a yellow oil: H NMR (400 MHz, CDCl₃): d=7.78 (d, J=8.0, 1H), 7.39

(d, J=1.6, 1H), 7.18 (dd, J=8.0, 1.6, 1H), 7.10 (s, 1H), 7.09 (s, 1H),

5.80 (d, J=1.2, 1H), 5.35 (d, J=1.2, 1H), 3.92 (s, 3H), 1.95 (s, 3H),

1.70 (s, 4H), 1.31 (s, 6H), 1.27 ppm (s, 6H); ¹³C NMR (100.6 MHz,

CDCl₃): d=165.9, 147.9, 145.7, 144.6, 142.4, 137.2, 133.9, 132.6,

131.5, 128.9, 128.2, 128.1, 128.0, 124.8, 117.7, 52.3, 35.1, 35.0, 33.9,

33.8, 31.9, 31.8, 19.9 ppm; IR (neat): n˜ =2955, 2924, 2860, 1734,

1714, 1598, 1542, 1497, 1456 cm^À1; GC–MS: m/z [M⁺] calcd for

C₂₅H₂₉O₂Cl: 396.1856, found: 396.1850.

2-Chloro-4-(1-(1,2,3,4-tetrahydro-1,1,4,4,6-pentamethylnaphtha-

len-7-yl)vinyl)benzoic acid (12): Compound 12 was synthesized

following the methods of Boehm and co-workers.^[11]To a 100 mL

round-bottom flask charged with 53 (0.353 g, 0.89 mmol) and

MeOH (5 mL) was added a 5m aq KOH (0.4 mL, 2.0 mmol). A reflux

condenser was fitted to the round-bottom flask and the reaction

solution was held at reflux and monitored by TLC. After 1 h at

reflux, the reaction solution was cooled to RT and quenched with

20% HCl (42 mL). The aqueous solution was extracted with EtOAc

(2ꢁ50 mL) and the organic extracts were combined, washed with

water and brine, dried over Na₂SO₄, and concentrated in vacuo to

give crude 12. Crude 12 was purified by column chromatography

(25 mL SiO₂, hexanes/EtOAc 9:1) to give 12 (0.31 g, 91%) as

1

a white crystalline solid, mp: 200–2018C: H NMR (400 MHz, CDCl₃):

d=7.98 (d, J=8.0, 1H), 7.43 (d, J=1.6, 1H), 7.21 (dd, J=8.0, 1.6,

1H), 7.11 (s, 1H), 7.09 (s, 1H), 5.83 (d, J=0.8, 1H), 5.38 (d, J=0.8,

1H), 1.96 (s, 3H), 1.71 (s, 4H), 1.31 (s, 6H), 1.28 ppm (s, 6H);

1

orless oil: H NMR (400 MHz, CDCl₃): d=10.40 (d, J=0.8, 1H), 8.33

(d, J=1.6, 1H), 8.10–8.08 (ddd, J=0.8, 1.6, 8.0, 1H), 7.95 (d, J=8.0

&10

&

www.chemmedchem.org

ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ChemMedChem 0000, 00, 1 – 17

ÝÝ

These are not the final page numbers!

RXR-Selective Agonists

MED

5H), 7.37–7.47 (m, 5H), 5.39 ppm (s, 2H); IR (neat): n˜ =2962, 1723,

Methyl 2-bromo-4-(1-(3,5,5,8,8-pentamethyl-5,6,7,8-tetrahydro-

naphthalen-2-yl)carbonyl)benzoate (50): Compound 50 was syn-

thesized according to the method of Boehm and co-workers.^[11]

Methyl 4-(chlorocarbonyl)-2-bromobenzoate (42) was synthesized

by refluxing 4-(methoxycarbonyl)-3-bromobenzoic acid (39) (1.64 g,

6.31 mmol) in SOCl₂(12.0 mL, 165 mmol) in a 100 mL one-neck

round-bottom flask fitted with a water-cooled reflux condenser.

Excess SOCl₂was removed in vacuo to give crude 42 as an off-

white solid, and this solid was dissolved in dry benzene (~20 mL)

and evaporated to dryness three times to remove residual SOCl₂.

The acid chloride 42 was dried on high vacuum to remove residual

benzene. To a 2-neck, 50 mL round-bottom flask equipped with

a reflux condenser and magnetic stir-bar was added 48 (1.38 g,

6.82 mmol) followed by a solution of crude acid chloride 42

(6.31 mmol) in CH₂Cl₂(15 mL). AlCl₃(2.20 g, 16.5 mmol) was added

to the reaction solution at RT slowly, with stirring, and the reaction

solution turned from colorless to red accompanied by the evolu-

tion of gas and heat. The reaction was stirred for 5 min then

heated at reflux for 15 min. The reaction was judged to be com-

plete by TLC, and the solution was poured into an ice solution

(25 mL) acidified with a 20% HCl solution (8 mL) and EtOAc was

added (13 mL). The aqueous and organic layers were separated,

and the aqueous layer was extracted with EtOAc (2ꢁ15 mL). The

combined organics were washed with water and brine, dried over

Na₂SO₄, filtered and concentrated to give crude 50. Crude 50 was

purified by column chromatography (250 mL SiO₂, hexanes/EtOAc

95:5 to 92.5:7.5) to give 50 (2.37 g, 84%) as a white, crystalline

1683, 1601, 1556, 1479 cm^À1

;

GC–MS: m/z [M⁺] calcd for

C₁₅H₁₁O₃Br: 317.9892, found: 317.9887.

4-((Benzyloxy)carbonyl)-2-bromobenzoic acid (33): Compound 33

was prepared according to the method of Kishida and co-work-

ers.^[32]To a 100 mL round-bottom flask charged with 30 (12.1 g,

37.9 mmol), sulfamic acid (4.06 g, 41.8 mmol), water (45 mL), and

ACN (32 mL) was added a solution of 80% NaClO₂(3.84 g,

42.5 mmol) in water (20 mL). After stirring for 1 h, the reaction so-

lution was poured into saturated Na₂SO₃(50 mL) and 1n HCl

(100 mL), and the resulting solution was extracted with EtOAc

(100 mL, thrice). The combined organic extracts were washed with

brine, dried over Na₂SO₄, and removed in vacuo to give crude 33

(10.5 g, 82%) that was used without further purification. A small

sample was purified by column chromatography (25 mL SiO₂, hex-

anes/EtOAc 1:1) to give pure 33 as a white powder: ¹H NMR

(400 MHz, CDCl₃): d=8.37 (d, J=1.6, 1H), 8.08 (dd, J=8.0, 1.6, 1H),

8.01 (d, J=8.0, 1H), 7.34–7.47 (m, 5H), 5.39 ppm (s, 2H); ¹³C NMR

(100.6 MHz, CDCl₃): d=170.3, 164.2, 135.6, 135.2, 134.4, 134.2,

132.1, 128.7, 128.6, 128.4, 128.2, 122.3, 67.6 ppm; IR (neat): n˜ =

2963, 2538, 1681, 1602, 1551, 1481 cm^À1; LC–APCI-MS: m/z [M⁺

+

H] calcd for C₁₅H₁₂O₄Br: 334.9919, found: 334.9934.

4-Benzyl 1-methyl 2-bromobenzene-1,4-dioate (36): Compound

36 was synthesized according to the method of Kishida and co-

workers.^[32]To a 100 mL round-bottom flask charged with com-

pound 33 (10.5 g, 31.3 mmol) was added SOCl₂(24.0 mL,

330 mmol) and the reaction solution was held at reflux for 1 h in

an oil bath at 848C. The reaction solution was cooled to RT and

the excess SOCl₂was removed in vacuo to give crude 4-chlorocar-

bonyl-2-bromobenzoic acid benzyl ester. The crude 4-chlorocar-

bonyl-2-bromobenzoic acid in dry toluene (16 mL) was added

dropwise to a solution of Et₃N (10.4 mL, 74 mmol) in MeOH

(106 mL, 2.6 mol) over 10 min with vigorous stirring. The reaction

solution was stirred 1 h and poured into 1n HCl (300 mL) and ex-

tracted with EtOAc (80 mL, thrice). The combined organic extracts

were washed with brine, dried over Na₂SO₄, and removed in vacuo

to give crude 36. Crude 36 was purified by column chromatogra-

phy (150 mL SiO₂, hexanes/EtOAc 95:5) to give pure 36 as an oil

(8.55 g, 78%): ¹H NMR (400 MHz, CDCl₃): d=8.32 (d, J=1.6, 1H),

8.03 (dd, J=8.0, 1.6, 1H), 7.80 (d, J=8.0, 1H), 7.36–7.46 (m, 5H),

5.38 (s, 2H), 3.95 ppm (s, 3H); ¹³C NMR (100.6 MHz, CDCl₃): d=

166.1, 164.3, 136.2, 135.3, 135.2, 133.7, 130.9, 128.7, 128.5, 128.4,

128.2, 121.4, 67.4, 52.7 ppm; IR (neat): n˜ =2964, 2558, 1723, 1688,

1603, 1557, 1484 cm^À1; GC–MS: m/z [M⁺] calcd for C₁₆H₁₃O₄Br:

347.9997, found: 348.0013.

1

solid, mp: 109–1118C: H NMR (400 MHz, CDCl₃): d=8.08 (d, J=1.6,

1H), 7.83 (d, J=8.0, 1H), 7.75 (dd, J=8.0, 1.6, 1H), 7.25 (s, 1H),

7.22 (s, 1H), 3.97 (s, 3H), 2.35 (s, 3H), 1.69 (s, 4H), 1.31 (s, 6H),

1.20 ppm (s, 6H); ¹³C NMR (100.6 MHz, CDCl₃): d=195.8, 166.2,

148.8, 142.0, 141.6, 135.6, 135.4, 134.8, 133.8, 130.9, 129.6, 128.7,

128.5, 121.5, 52.7, 34.8, 34.7, 34.3, 33.8, 31.6, 31.5, 20.0 ppm; IR

(neat): n˜ =3015, 2953, 2931, 2863, 1736, 1666, 1605, 1544, 1496,

1459 cm^À1; GC–MS: m/z [M⁺] calcd for C₂₄H₂₇O₃Br: 442.1144, found:

442.1135.

Methyl 2-bromo-4-(1-(1,2,3,4-tetrahydro-1,1,4,4,6-pentamethyl-

naphthalen-7-yl)vinyl)benzoate (54): Compound 54 was synthe-

sized according to the method of Boehm and co-workers.^[11]To

a 20-dram vial containing 50 (0.923 g, 2.08 mmol) and dry THF