Development of Subtype-Selective Oestrogen Receptor-Based Therapeutics

REVIEWS

Development of subtype-selective oestrogen receptor-based therapeutics

Stefan Nilsson*‡, Konrad F. Koehler* and Jan-Åke Gustafsson‡§ Abstract | The two oestrogen receptor subtypes α and β are hormone-regulated modulators of intracellular signalling and gene expression. Regulation of oestrogen receptor activity is crucial not only for development and homeostasis but also for the treatment of various diseases and symptoms. Classical selective oestrogen receptor modulators are well established in the treatment of breast cancer and osteoporosis, but emerging data suggest that the development of subtype-selective ligands that specifically target either oestrogen receptor-α or oestrogen receptor-β could be a more optimal approach for the treatment of cancer, cardiovascular disease, multiple sclerosis and Alzheimer’s disease.

Detrusor muscle Oestrogen receptors (ERs) are intracellular transcription For example, the expression of ERα but not ERβ fluctuates Bladder muscle that relaxes to factors whose activity is modulated by the naturally depending on the stage of rat mammary gland develop- fill the bladder with urine or occurring oestrogens in the body or by synthetic, ment, going from relatively high levels at prepubertal stage contracts to squeeze urine out non-steroidal, non-hormonal agonist and antagonist to very low levels at pregnancy and then back up to high from the bladder. ligands. Oestrone (E1), oestradiol (E2; also known as levels during lactation, followed by a dramatic decrease 17β‑oestradiol) and oestriol (E3) are the three main at the postlactation stage4. During tumour progression endogenous oestrogens, of which E2 is the predomi- in invasive breast carcinoma, the expression of ERβ nant and most biologically active. E2 is produced from decreases compared with normal mammary tissue5, and testosterone by the aromatase enzyme cytochrome P450 oestrogen deficiency in the female rat leads to degenera- 19A1 (REF. 1). The primary sites of E2 production are the tion of the bladder ultra structure owing to an increased ovaries in women of fertile age and extragonadal tissues ratio of ERα to ERβ in the detrusor muscle; however, E2 (adipose tissue, brain, vascular endothelial cells, bone supplementation resulted in a significant increase in ERβ cells, breast tissue and skin fibroblasts) in postmeno expression and bladder function6. Whole-body glucose pausal women. In men, E2 is produced by the testes and intolerance induced by high-fat diet in female rats was extragonadal tissues. The physiological role of oestro- associated with decreased ERα expression in adipose gens includes sexual maturation and fertility, regulation tissue7, and in the mouse central nervous system, the of lipid and carbohydrate metabolism, skeletal develop- level of cytoplasmic ERα or ERβ in the dorsal hippocam- ment and integrity, and homeostasis of the cardiovascular pal region varied with the oestrous cycle8. Furthermore, *Karo Bio AB, Novum, and central nervous systems. in synapses of the rat hippocampal CA1 region, both SE‑141 57 Huddinge, Sweden. Today, we are aware of the existence of two ER ERα and ERβ decreased with age but, in contrast to ‡ Department of Biosciences subtypes; ERα, which is encoded by the ESR1 gene on ERα, the expression of ERβ increased in response to E2 and Nutrition, Karolinska 9 Institutet, Novum, chromosome 6, and ERβ, which is encoded by the ESR2 in older animals . SE‑141 83 Huddinge, gene on chromosome 14. Both subtypes are expressed Phenotypic characterization of knockout mice defi- Sweden. in a wide range of tissues and cell types throughout the cient in either the expression of ERα or ERβ has revealed §Center for Nuclear body; ERα has been found predominantly in bone, male interesting differences between the physiological roles Receptors and Cell Signalling, Department of Biology and reproductive organs (testes and epididymis), prostate of the two receptor subtypes (Supplementary informa- Biochemistry, University (stroma), uterus, ovary (thecal cells), liver, mammary tion S1 (table)). ERα has a more profound effect on of Houston, Houston, gland, adipose tissue, heart, vascular system and brain. the development and function of the mammary gland Texas 77204‑5056, USA. ERβ is mainly present in the bladder, prostate (epithelium), and uterus and on the maintenance of metabolic and Correspondence to J.-Å.G. ovary (granulosa cells), colon, adipose tissue, immune skeletal homeostasis. ERβ, however, has more pro- e‑mail: [email protected] 2,3 doi:10.1038/nrd3551 system, heart, vascular system, lung and brain . The nounced effects on the central nervous system and on Published online presence of ERα or ERβ has been shown to vary in differ conditions of cellular hyperproliferation. Both receptors 16 September 2011 ent tissues during development, ageing or disease state. have important roles in the development and function

Progestagen of the ovaries and in the protection of the cardiovascular fact decreases cardiovascular disease risk in women if Collective name for a ligand of system. More knowledge of the nature of ER subtype- MHT is initiated soon after menopause or before the the progesterone receptor, selective ligands (whether they are agonists or antago- age of 60 years11–13. The WHI results and conclusions such as progestogen, gestagen nists) and the biological roles of ERα or ERβ in different caused, however, a dramatic drop in the prescriptions for or progesterone. tissues could be used to decide which ER subtype to and use of MHT in the United States14 and worldwide15 target for the optimal treatment of various diseases and and an increased interest in the development of ER symptoms. E2 and classical selective oestrogen receptor subtype-selective modulators because of their poten- modulators (SERMs) target both of the ER subtypes, but tial to retain the benefits of MHT without the severe targeting just one subtype may lead to a more efficacious adverse events. therapy with less risk of side effects (BOX 1). ERα and ERβ regulate different classes of genes by A decline in oestrogen production in women as a binding to different sites on DNA and by the recruitment result of ceasing ovarian function during menopausal of different co-regulatory and chromatin remodelling transition substantially increases the risk of osteoporotic proteins16. ERβ-selective agonists can be divided into fractures and coronary heart disease (CHD), and can two major classes: those that are affinity and potency cause symptoms such as hot flushes and vaginal atrophy. selective for ERβ and those that are ERβ efficacy selec- During the 1990s, menopausal hormone therapy (MHT) tive16. ERβ-selective agonists do not stimulate the pro- — which primarily consisted of conjugated equine oes- liferation of breast or endometrial tissues. By contrast, trogens (CEEs) in combination with medroxyprogester- activation of ERα with the ERα-selective agonist pro- one acetate (MPA; a progestagen to prevent the risk for pyl pyrazole triol (PPT) is associated with proliferative endometrial cancer development) — was frequently responses in breast and uterine tissues17,18. Thus, selec- prescribed for treatment of menopausal symptoms, tive agonist activation of ERα imposes increased risk for prevention of postmenopausal bone loss and bone frac- the development of breast or endometrial cancer, which ture risks, and prevention of the development of CHD. makes the design and synthesis of ERα-selective agonists However, a study published by the Women’s Health a less-attractive approach for development of novel thera- Initiative (WHI) stated that combined oestrogen and pies. However, development of ERα-selective antagonists progestagen therapy should not be initiated or continued or ERα-selective ligands with mixed tissue selective ago- for the prevention of CHD owing to lack of efficacy to nist/antagonist activity (ERα-selective SERMs) may hold prevent cardiovascular events and increased risk of breast therapeutic promise (BOX 1). cancer10. It was stated that the overall health risks with In this article, we review the roles of ERα and ERβ in combined MHT exceeded its benefits10. The conclusions various animal disease models with a focus on the thera drawn from the WHI study have been widely discussed peutic opportunities of agonists that selectively target and criticized; the beneficial or detrimental effects of the ERβ subtype for treatment of cancer, cardiovascular MHT on cardiovascular outcome have been suggested disease, multiple sclerosis and Alzheimer’s disease. to depend on timing. Follow-up studies in women aged between 50 and 59 clearly showed that MHT in ER signalling and mechanism of action ER architecture. Both ER subtypes have a modular domain structure that is common to all members of Box 1 | SERMs and ER subtype-selective modulators the nuclear hormone receptor family19,20 (FIG. 1a). This domain structure consists of an amino-terminal activa- The acronym SERM (selective oestrogen receptor modulator) was coined in the early tion function 1 (AF1) domain for ligand-independent 1990s, before knowledge of the existence of two oestrogen receptor (ER) subtypes, and so it refers to a set of anti-oestrogenic ligands (for example, tamoxifen and raloxifene) activation of transcription, a central DNA binding with mixed, tissue-selective agonist and antagonist activity. This type of ligand displays domain (DBD) for sequence-specific binding to DNA, oestrogen antagonism in breast and uterine tissues, but more oestrogen-like activity in and a carboxy‑terminal AF2 domain with a multifunc- tissues such as bone and liver130. Non-subtype-selective SERMs (classical SERMs) are tional ligand binding domain (LBD). The LBD contains used in the clinic for treatment of ERα-positive breast cancer and/or for the prevention a ligand-binding cavity and is responsible for ER dimeri- and treatment of postmenopausal osteoporosis130. However, SERMs have been reported zation and ligand-dependent activation of transcription. to increase the risk for thromboembolic events and stroke, and tamoxifen is known to 131 increase the risk for endometrial cancer and cataract . Moreover, owing to their Nuclear activity of ERs. Nuclear (genomic) modulation anti-oestrogenic nature, they may also increase the incidence of hot flushes and of target gene transcription by the ERs is a complex and affective or emotional disorders. Two SERMs (levormeloxifene and idoxifene) were also highly dynamic process. After natural or synthetic ligands reported to increase the incidence of uterine prolapse132,133. The tissue-selective mixed agonist and antagonist activity of classical SERMs is bind to the LBD, the ER dissociates from its heat shock 21 dependent on the ER subtype expressed, the cell’s dependence on the two chaperone proteins , undergoes a conformational change transactivation functions of the ERs, the gene promoter context and the distinct and dimerizes. The receptor dimer then binds to target receptor conformation induced by each SERM that dictates the specific interactions sites on DNA either directly or indirectly (through other between the liganded ER and proteins such as co-activators and co-repressors in the transcription factors such as activator protein 1 (AP1) target cell20,130,134–136. and specificity protein 1 (SP1)22). After binding, the ER modulators that are ER subtype-selective — that is, ligands that preferentially bind dimer interacts with co-regulatory proteins (chromatin to and activate one or the other ER subtype (ERα or ERβ) — are also often called SERMs. modulators, co-activators and basal transcription fac- This may be confusing, as the nature and biology of classical SERMs (for example, tors), which are required for the efficient modulation tamoxifen, raloxifene and others137) is radically different. Classical SERMs act through of gene expression by ERs23 (FIG. 1b). Gene regulation by both ERα and ERβ with similar potency but with different efficacy138. the ERs has been demonstrated to be a cyclic process;

C ERs promote several rounds of transcription initiation in '4α #( &$& #( response to an agonist (for example, E2), and each round of initiation is followed by release of the ER transcription '4β #( &$& #( complex from the promoter. If the agonist remains present, liganded ER binds again to DNA to initiate another D 'ZVTCEGNNWNCTURCEG '4NKICPF round of transcription activation. This cyclic mode of gene regulation allows the cell to sense and adapt to fluc- tuations in the level and nature of the ligand present24. The nature and concentration of ligand present deter- mines whether homodimers (ERα–ERα or ERβ–ERβ) or %[VQRNCUO heterodimers (ERα–ERβ) are formed, or a combination of the two25. Ligands that selectively bind ERβ promote 0WENGWU the formation of ERβ homodimers only. By contrast, ligands that are selective for ERα stimulate the formation %QTGIWNCVQTU %QTGIWNCVQTU %QTGIWNCVQTU of both ERα homodimers and ERα–ERβ heterodimers, suggesting that ERα has a dominant role in the formation of heterodimers25. Ligand-dependent homo- or hetero #2 52 dimer formation has been shown to result in distinct patterns of gene regulation26, which has implications for **7&$1117*$&& the scope of therapy with ERs and the risk of side effects. )TQYVJHCEVQT )TQYVJ Ligands with significantly higher ERβ affinity relative HCEVQT E 01 to ERα would minimize the risk of side effects (such G015 TGEGRVQT %CXGQNC as breast and endometrial cancer) mediated by activa- 5VTKCVK 2 tion of ERα. A genome-wide study of the occupancy of '4 P 2+-s#-6 54% '4 R ERα and ERβ, respectively, to binding sites on chroma- )α (#- 27 K '4 tin unmasked further complexity . When ERα or ERβ /#2- 54% 54% /#2- is present alone in a cell, there is a substantial overlap #-6 /#2- #-6 in their chromatin binding sites in the presence of E2. 2+-s#-6 ,0- When present together, however, fewer sites were shared, %C 2 2 as each ER subtype restricts the other one for binding site #EVKXCVKQPQHE[VQRNCUOKEVCTIGVU '4 selection, with ERα being more dominant in competing for binding sites than ERβ. Oestrogen response element 2QUVVTCPUNCVKQPCNOQFKȮECVKQPQHEQCEVKXCVQTU CPFQTEJTQOCVKPTGOQFGNNKPIHCEVQTU EQTGIWNCVQTU (ERE) binding sites were preferred by both ER subtypes when present alone in the cell, but when present 2 2 2 2 together, ERα displaced ERβ into new binding sites that %QTGIWNCVQTU %QTGIWNCVQTU were less enriched for EREs but more enriched for other transcription factor DNA binding motifs. In the pres- 2 2 6TCPUETKRVKQP ence of the ERα-selective agonist PPT, the pattern of ERα HCEVQT chromatin binding sites was similar to the pattern of binding sites in the presence of E2. Furthermore, the bind- Figure 1 | Architecture of oestrogen receptors and their mode of nuclear and ing pattern remained the same if ERα was present alone cytoplasmic activity. a | Both receptors have a domain0CVWT structureG4G XKGYUthat is^ common&TWI&KUEQ to XGT[ or was together with ERβ in the cell. By contrast, the other nuclear receptors, including an amino‑terminal activation function 1 (AF1) domain presence of the ERβ-selective agonist ERB‑041 resulted that is involved in ligand-independent activation of transcription, a central DNA binding in a pattern of ERβ chromatin binding sites that was dis- domain (DBD) that recognizes and binds to specific sequences on DNA, and a tinct from the pattern of binding sites in the presence of carboxy-terminal AF2 domain containing a multifunctional ligand binding domain (LBD) that is involved in receptor dimerization, ligand binding and ligand-dependent E2 in cells containing ERβ alone; this suggests that the activation of transcription. b | Oestrogen receptor (ER) ligands are thought to enter the difference in ERβ conformation induced by ERB‑041 cell by passive diffusion through the cellular membrane and to bind ERs located in the compared with E2 influences ERβ chromatin binding cytoplasm. Following binding, the receptor protein undergoes a conformational change, site occupancy27. translocates to the nucleus and dimerizes. In the nucleus, the ER binds either directly to Agonist ligands induce a conformation in the LBDs its specific oestrogen response element (ERE) sequences on DNA (the canonical ERE of ERs that allows AF2 to form a stable interaction sequence is displayed on the lower left) or by tethering onto other transcription factors, with co-activator proteins through the nuclear receptor such as activator protein 1 (AP1) and specificity protein 1 (SP1). For efficient modulation (NR)‑box motif (LxxLL, in which ‘x’ indicates any amino of gene expression, the ER interacts with various co-regulatory protein complexes, acid)20,23,28. It has been demonstrated that there is both an which then signal to the gene transcription machinery. c | Schematic representation of ER subtype-selective as well as a ligand-specific recruit- various cytoplasmic signalling pathways mediated by ligand-bound ERs localized in the 29,30 cytoplasm, at the plasma membrane or in caveolae. Rapid cytoplasmic signalling may ment of NR‑boxes to the receptor , suggesting that result in the activation of cytoplasmic targets or integrate with nuclear regulation of different agonists induce a distinct receptor conforma- 16,31 gene expression. Growth factors acting through their receptors at the cellular membrane tion when bound to ERα or ERβ . This ligand-specific may also activate ERs through phosphorylation of the ER. eNOS, endothelial nitric modulation of receptor conformation and subsequent oxide synthase; FAK, focal adhesion kinase; JNK, c‑Jun N‑terminal kinase; MAPK, recruitment of a specific co-regulatory complex was mitogen-activated protein kinase; NO, nitric oxide; PI3K, phosphoinositide 3‑kinase. shown to have a dramatic effect on the ability of ERβ

to inhibit inflammatory response genes31. Thus, the Almost 1,300 genes were regulated in response to E2, preferences of co-activators for specific ligand–ER and of those, 243 genes were regulated by both E2 and complexes and the subsequent consequences of co- EDC. Ninety-two genes were considered to be uniquely activator recruitment on target gene regulation may regulated by EDC. None of the genes stimulated by EDC, provide new, unforeseen therapeutic opportunities31. however, resulted in recruitment of ER to target sites In contrast to oestrogen agonists, oestrogen antagonists on DNA, suggesting that other downstream effectors of induce a receptor conformation that leads to association EDC-triggered cytoplasmic signalling cascades may be with co-repressor proteins — such as nuclear receptor responsible for the EDC-mediated gene regulation, such co-repressor 1 (NCOR1) and silencing mediator of as phosphorylation of co-activators and their associa- retinoic acid and thyroid hormone receptor (SMRT; also tion with transcription complexes or phosphorylation of known as NCOR2) — thereby shutting off target gene chromatin remodelling factors39,41,42. transcription20,23,28. Thus, the nature of the ligand is a key Rapid ER‑mediated cytoplasmic signalling also takes determinant of co-regulatory protein recruitment and of place in neuronal cells. In human and mouse embryonic the distinct biological responses to ER ligands. The devel- stem cell-derived neurons, E2 and the selective ERβ ago- opment of ERα-selective antagonists with mixed tissue nist diarylpropionitrile (DPN), but not PPT, increased selective agonist/antagonist activity could be a thera- intracellular Ca2+ within a few minutes, and this effect peutic opportunity for the treatment of breast cancer was dependent on multiple cytoplasmic kinase signal- and osteoporosis, and such treatments could replace the ling pathways, including the MAPK and PI3K–AKT existing classical, non-subtype-selective SERMs (BOX 1). pathways43. In rat hippocampal neurons, E2 promotes However, owing to a lack of data, it would be difficult neuronal survival by increasing intracellular Ca2+ and to predict the risk–benefit of ERβ-selective antagonists activating MAPK signalling. This effect of E2 can be with mixed tissue selective agonist/antagonist activity. mediated by both ERα and ERβ44. All ER ligands probably have both nuclear and cyto- Rapid cytoplasmic ER signalling. In addition to directly plasmic effects. However, there may be applications for modulating gene transcription activity in the nucleus, drugs that act only at cytoplasmic ERs; for example, the ERs are involved in rapid (from seconds to minutes) activation of cytoplasmic ERα stimulates repair of the cytoplasmic stimulation of protein kinase pathways, vascular endothelial cell layer following injury36. Cyto phospholipase C activation, mobilization of intracellular plasmicactivation of ERα did not promote breast cancer calcium, modulation of potassium currents and increases growth or uterotrophic activity in vivo as compared to in nitric oxide production, among other effects32,33. E2, perhaps suggesting a therapeutic opportunity for The mechanism by which the ERs become associated ERα-selective agonists with exclusive cytoplasmic activity with the cellular plasma membrane is not fully under- (for example, a PPT–dendrimer conjugate). The pros and stood, but palmitoylation of ERα and ERβ is one mecha- cons of activating ERβ solely in the cytoplasm have not nism that has been shown to allow the ERs to associate yet been studied and are therefore difficult to assess. with caveolae rafts at the cellular membrane34 (FIG. 1c). Direct interaction of the ERs with the caveolin‑1‑binding Development of ER subtype-selective ligands protein striatin35 in vascular endothelial cells was found ER pharmacophore. The basic requirements for high- to be required for the rapid activation of protein kinase affinity ligand binding to ERα are well known owing to

pathways by E2, formation of the ER–Gαi complex and an extensive structure–activity relationship (SAR) that activation of endothelial nitric oxide synthase (eNOS) has been developed over the past 50 years45. The ER (FIG. 1c), which together results in stimulation of carotid pharmacophore consists of two hydroxyl groups that are re-endothelialization and attenuation of the development ~11 Å apart and are joined by a hydrophobic core (FIG. 2). of vascular wall hyperplasia35–37. In addition, at least one of the two hydroxyl groups in In the human breast cancer cell line MCF‑7, methyla- preferred ligands is bonded to an aromatic ring to form tion of Arg260 in the ERα DBD was shown to trigger a phenolic substructure. the formation of a cytoplasmic complex containing ERα, The elucidation of the structure of ERα LBD46 pro- Palmitoylation SRC, focal adhesion kinase (FAK; also known as FADK1) vided an explanation for this SAR: the amino acid resi- The covalent attachment of and the p85 subunit of phosphoinositide 3‑kinase (PI3K) dues Glu353 and Arg394 that line the ‘A-ring pocket’ fatty acids, such as palmitic in response to E2 (REF. 38). This macromolecular com- of the receptor form strong hydrogen bonds with the acid, to cysteine residues on proteins, which facilitates plex then signals downstream to the serine/threonine phenolic hydroxyl group of E2, while the 17β‑hydroxyl the interaction of the protein kinase AKT pathway, leading to cell proliferation and group of the steroidal hormone is hydrogen bonded with cell membranes. cell survival38 (FIG. 1c). to His524 in the ‘D-ring pocket’. The remainder of the Activation of both nuclear and cytoplasmic ER sig- ligand binding cavity is lined by hydrophobic amino acid Vascular wall hyperplasia Excessive proliferation nalling has been shown to be integrated in the regulation side chains that are compatible with the hydrophobic 39 and growth of cells in the of gene expression in MCF‑7 cells . The action of the scaffold of E2. Most ligands contain a phenol or phenol vascular wall. oestrogen–dendrimer conjugate (EDC), which exclusively bioisostere that binds to the A‑ring pocket. The second activates cytoplasmic ER signalling, was compared to hydroxyl group is less important for affinity and can be Dendrimer the activity of E2 in MCF‑7 cells39,40. Although both E2 dispensed with or replaced by a hydrogen bond accep- A highly branched and roughly spherical polymeric compound and EDC rapidly activated MAPKs (mitogen-activated tor that can either bind directly to His524 in the D‑ring that is formed by attachment protein kinases), SRC and c‑Jun N‑terminal kinases, only pocket or indirectly to His524 through a bridging water of monomers to a central core. E2 also triggered ER‑mediated nuclear events (FIG. 1c). molecule45.

C βHCEG E )NW ¡ .GW OH .GW .GW /GV /GV H OH .GW *KU *KU H H H%& HO )NW # $ H H HO +NG *[FTQRJQDKEEQTG #TI 2JG 2JG .GW /GV β8' αHCEG /GV +NG D /GV .GW '4βDKPFKPIECXKV[ #TI

βHCEG '4αDKPFKPIECXKV[ F )NW 6GVTCJ[FTQȯWQTGPQPG )NW *KU )NW *KU

*KU

O O #TI H #TI

H O O /GV /GV O H +NG )GPKUVGKP αHCEG +NG #TI 2JG PQVUJQYP Figure 2 | Oestrogen receptor binding cavity and ligand subtype selectivity. a | Schematic diagrams of the oestrogen receptor (ER) pharmacophore (left) and the ERβ binding cavity (right). The endogenous ligand oestradiol is used to define the locations of the ligand-binding pocket within the receptor in analogy to the Schechter0CVWT and G4GBergerXKGYU nomenclature^&TWI&KUEQ thatXGT[ has been developed for proteases and their substrates139. Stereogenic substituents that are above the plane of the steroid are referred to as β, whereas substituents below the plane are referred to as α. In analogy to the steroid nomenclature, the region of the binding cavity above the plane of the bound hormone is referred to as the β-face, whereas the area below the plane is referred to as the α-face. Furthermore, regions of the receptor that are closest to the A-, B-, C- and D‑rings of the bound receptor are referred to as A-, B-, C- and D‑ring pockets, respectively. b | Origin of ERβ selectivity of flat ligands. Depicted is a superposition of crystallographic structures of ERα (light blue = carbon, red = oxygen, dark blue = nitrogen, yellow = sulphur) and ERβ (green = carbon), both complexed with the ERβ-selective ligand genistein (Protein Data Bank identifiers: ERα–genistein = 1X7R; ERβ–genistein = 1X7J). Portions of the ERα binding cavity are shown as solid surfaces, while the corresponding regions in ERβ are displayed as dot surfaces. The two surfaces in ERβ approach the ligand more closely compared to the surfaces in ERα. The net result is that the binding cavity of ERβ is slightly narrower compared to the binding cavity of ERα, and the narrower cavity allows more energetically favourable packing of flat ligands. c | Comparison of the binding cavity of ERα versus the binding cavity of ERβ when in complex with the ERβ-selective ligand 8β‑vinyl-oestradiol (8β-VE2; grey), where the receptor diagrams are overlaid on top of each other57. The side chains of the two amino acid residues that differ between ERα (orange) and ERβ (magenta) and important residues that engage in hydrogen bonding (yellow dashes) are displayed as sticks. A steric repulsion between the vinyl group of the ligand and the C5 methyl group of the branched side chain of Leu384 substantially reduces affinity for ERα. This steric repulsion does not exist in ERβ, in which the branched Leu384 is replaced by the linear Met336 side chain. Where two residue numbers are separated by a solidus, the first number refers to the residue in ERα and the second number refers to the comparable residue in ERβ. d | Comparison of the binding cavity of ERα versus the binding cavity of ERβ when in complex with the ERβ-selective ligand tetrahydrofluorenone58 (FIG. 3c), where the receptor diagrams are overlaid on top of each other. The butyl side chain of the ligand displaces Phe377, opening up a binding pocket that does not exist in the absence of ligand. In ERα, the longer Met421 side chain (magenta) constricts the width of the pocket, resulting in a repulsive steric clash with the butyl side chain of the ligand. In ERβ, Met421 is replaced with the shorter, branched Ile373 (orange), resulting in a wider pocket that easily accommodates the side chain of the ligand.

After the discovery of ERβ, subsequent studies as research tools and potential therapeutics for a wide showed intriguing differences in the tissue distribu- range of conditions47,48. Given the moderate sequence tion of and gene regulation by ERβ relative to ERα, and and high structural conservation between the LBDs of since then there has been an intense effort by academic the two ER subtypes, it is not surprising that the result- and industrial groups to develop ERβ-selective ligands ing SAR for ERβ has been found to closely recapitulate

(TABLE 1) Phyto-oestrogens that of ERα. Nevertheless, there are subtle differences and ERβ . The first is located in the β-face of Plant-derived heterocycles, between the two ligand binding pockets that have been the binding cavity where Leu384 in ERα is replaced primarily isoflavones, exploited to develop subtype-selective ligands. by Met336 in ERβ. The second is located in the α-face that mimic the activity where Met421 in ERα is replaced by Ile373 in ERβ. of endogenous oestrogens by binding to oestrogen Types of selectivity. The most obvious way of obtaining Both of these amino acid differences are considered receptors. subtype-selective ligands is to maximize the differences to be conservative substitutions in that all four of the in binding affinity between the two receptor subtypes. amino acids involved occupy similar volumes and have Transrepression It was determined shortly after the discovery of ERβ similar hydrophobicities (Supplementary information The process whereby a that various phyto-oestrogens such as genistein display S2 (table)). Although the volumes occupied by these regulatory protein — for 49 example, an oestrogen a modest binding selectivity for ERβ relative to ERα . amino acid residues are very similar, the shapes of the 50 receptor — inhibits or Synthetic ligands, such as DPN , were then discovered amino acid side chains differ. Furthermore, owing to represses the activity of that also displayed modest binding selectivity for ERβ. the lower rotational barrier of the methione side chain another regulatory protein However, there are additional mechanisms that can be compared to isoleucine or leucine53, the α-face of the through protein–protein interaction. exploited to obtain selectivity. Ligand binding is a neces- ERα binding cavity and the β-face of the ERβ cavity sary but insufficient condition for regulating the activity are ‘softer’ than the β- and α-surfaces in ERα and ERβ, of the receptor. The ligand must also change the confor- respectively. Consequently, ligand substituents that are mation of the receptor in such a way that the receptor’s oriented downwards pointing in the direction of Met421 affinity for co-regulatory proteins is altered51. The result- (for example, the lactone ring of 16α‑lactone oestradiol ing changes in the composition of ER transcriptional (16α‑LE2)) tend to be better accommodated in ERα and complexes either enhance or decrease gene expression. contribute to binding selectivity for ERα, whereas those For example, the genes that are upregulated by ERβ when that are pointed upwards facing Met336 (for example the it is activated by the natural product liquiritigenin only vinyl group of 8β‑vinyl-oestradiol (8β‑VE2; Schering)) partially overlap with the genes that are upregulated when generally prefer binding to ERβ. There is one additional ERβ is activated by E2 (REF. 52). Furthermore, the pat- difference in the ‘second shell’ of amino acid residues tern of differentially regulated genes is cell-type specific. that line the binding cavity. This difference is roughly in Several dozen genes were found to be upregulated by the plane of the bound hormone in the B‑ring pocket, liquiritigenin that were not upregulated by E2 in human where Val392 in ERα is replaced by Met344 in ERβ. U2OS osteosarcoma cells52. This raises the possibility that Given the limited number and conservative nature of ERβ-selective ligands may display unique pharmacology the amino acid changes between ERα and ERβ that line above and beyond that caused by their subtype selectivity. the binding cavity, it has been a daunting challenge to In addition to directly regulating the expression of develop subtype-selective ligands. genes through binding to EREs (that is, transactivation), Hydrogen bonding interactions between the hydroxyl ERβ can repress the activity of other transcription factors groups of oestrogenic ligands and their receptors are to such as AP1 in a manner similar to the glucocorticoid amino acid residues that are conserved between the receptor, resulting in an anti-inflammatory response31. α- and β-subtypes (Glu353, Arg394 and His524 in ERα, ERβ-selective agonists such as 5‑androstene‑3β,17β‑diol and Glu305, Arg346 and His475 in ERβ); therefore, such and chloroindazole (FIG. 3b) efficiently transrepress and interactions do not contribute directly to receptor subtype prevent experimental autoimmune encephalomyelitis selectivity. However, these hydrogen bonds do contribute (EAE), whereas the ERβ-selective agonist DPN and the to selectivity against more distantly related receptors, such non-selective agonist E2 are unable to signal through as the androgen, glucocorticoid, mineralocorticoid and this pathway31. progesterone receptors46. In addition, these hydrogen bonds restrict the binding modes available to ligands, Exploiting differences between ERα and ERβ. A com- making it possible to more precisely position ligand parison of the ligand binding cavities reveals only two functional groups adjacent to the pair of amino acid conservative amino acid differences between ERα residues that differ between ERα and ERβ and thereby enhance subtype selectivity.

Table 1 | Structural features that confer ER subtype selectivity The origin of selectivity of flat ligands. The first ERβ- Feature Binding site residues involved Subtype-selective ligands selective ligands to be discovered were phyto-oestrogens (such as the heteroaromatic genistein), which possesses ER ER ER ER α β α β a largely planar topology. Many of the ERβ-selective Cavity Leu384, Met336, None known Genistein49 ligands that have been developed since then retain simi- height Met421 Ile373 and other lar planar topologies (FIG. 3). The explanation for the ‘flat’ ligands selectivity of these flat ligands is that the binding cavity β -face Leu384 Met336 None known 8β-VE2 (REF. 57) of ERβ is slightly narrower than that of ERα54 (FIG. 2b). 58 α-face Met421 Ile373 PPT, 16α-LE2 Fluorenone This allows better packing of flat aromatic ligands such (REF. 57) as genistein in ERβ compared with ERα55. In particular, B-ring Val392 Met344 None known 4-benzyl the greater flexibility of Met336 in ERβ compared with cavity chromanol60 Leu384 in ERα allows a more favourable interaction 54 16α-LE2, 16α-lactone oestradiol; 8β-VE2, 8β-vinyl-oestradiol; ER, oestrogen receptor; between Met336 and the central B‑ring of genistein . PPT, propyl pyrazole triol. This ‘snug-fit’ mechanism of selectivity has been

C '4αUGNGEVKXGNKICPFU OH O HO CH3 O 2TQR[NR[TC\QNGVTKQN 226 H H N OH N HO α.' 5EJGTKPI HO

D '4βUGNGEVKXGNKICPFURNCPCT O O OH Cl OH N OH N O OH O OH Cl HO HO HO HO )GPKUVGKP .KSWKTKVKIGPKP 91 %JNQTQKPFC\QNG DGP\QȯWQTGPG#M\Q0QDGN N OH C OH OH N O OH O

C HO HO HO N HO F F &KCT[NRTQRKQPKVTKNG #75 5 'SWQN '4$ '4$ &20 #WUKQ2JCTOC RTKPCDGTGN9[GVJ PCRJVJQPKVTKNG9[GVJ

E '4βUGNGEVKXGNKICPFUPQPRNCPCT OH OH O OH Br O CH3 H2C CH3 O HO

H H HO HO CH3 HO β8' 5EJGTKPI 91 91 'TVGDGTGN EJTQOCPQN#M\Q0QDGN VGVTCJ[FTQȯWQTGPQPG EJTQOGPQN'NK.KNN[%Q /GTEM%Q

F 5WRGTRQUKVKQPU 2NCPCT 0QPRNCPCT

6QR 6QR

5KFG 5KFG

Figure 3 | Oestrogen receptor ligands. a | ERα-selective ligands. b | Planar ERβ-selective ligands. AUS‑131 (S)-Equol is in Phase II clinical trials for the treatment of benign prostate. ERB‑041 was discontinued after Phase II trials for rheumatoid arthritis and ERB‑196 was discontinued after Phase I trials for inflammatory bowel disease and sepsis. c | Non-planar 0CVWTG4GXKGYU^&TWI&KUEQXGT[ ERβ-selective ligands. d | Superposition of the receptor-bound conformations of the eight planar and four non-planar ligands from b and c. For each class, orthogonal views (top and side) are depicted. A comparison of the two side views demonstrates that the molecular volumes of the planar ligands are largely confined to a single plane, whereas each of the non-planar ligands possesses at least one substituent that is approximately perpendicular to the main plane of the ligand. The receptor-bound conformations were determined either from an experimental crystallographic structure of the ligand complexed with ERβ (Protein Data Bank (PDB) identifiers: 1X7J (genistein), 1X7B (ERB-041; also known as prinaberel), 1YYE (ERB-196), 2GIU (tetrahydrofluorenone) and 2I0G (erteberel)) or from a modelled structure in which the ligand was docked and energy minimized within an ERβ crystal complexed with a close analogue of the ligand (PDB identifiers: 1X7J (genistein; used as a template to model the rest of the structures shown in b and c) and 3OLS (oestradiol; used as a template to model 8β-vinyl-oestradiol (8β-VE2) and WO-01064665)). 16α-LE2, 16α lactone oestradiol.

suggested to explain the selectivity of potassium chan- van der Waals distance of the two interacting partners) nels for potassium cations over the smaller sodium and the interaction energy at longer distances is small, cations56. However, this type of selectivity has its leading to modest selectivities. limits, as it follows the Lennard–Jones potential energy curve (Supplementary information S3 (figure)) that Escaping flatland. At separation distances between the shows weak attraction at the energy minimum of the ligand and protein that are shorter than the optimum, curve and less attraction at longer distances. The energy strong repulsive interactions rapidly develop. Ligands difference between optimal separation (the sum of the that enforce strong repulsive interactions with ERα while

avoiding corresponding unfavourable interactions with agar (an in vitro indication of in vivo tumorigenic poten- ERβ tend to display much greater and more robust tial). By contrast, silencing of ERα led to apoptosis in the selectivity compared to ligands that achieve selectivity presence of E2, in this case mediated by ERβ18. Moreover, through differences in attractive interactions. To selective ERα activation with PPT resulted in stimula- achieve robust selectivity, the ligand must display three tion of cell proliferation, whereas selective activation characteristics. The first requirement is that the ligand of ERβ with DPN inhibited cell growth and decreased must position substituents very close to one or more the cell number by inducing cell death18. In the human of the amino acid differences between ERα and ERβ breast cancer cell line MCF‑7, the presence of ERα alone, in order to develop strong repulsive interactions to E2 or a high concentration of genistein stimulated the ERα. However, maintaining this repulsive interaction expression of genes that promote cell cycle progression, with ERα may be difficult because of the possibility of whereas the presence of ERα and ERβ resulted in the alternative binding modes and conformational flex- stimulation of genes involved in negative regulation of ibility in both the ligand and the receptor. Hence there is proliferation and attenuation of ERα-upregulated pro- an additional requirement that the ligand be confor- liferative genes67. In ERα-expressing T47D cells, more mationally rigid so that it is less likely to find alter- than 1,400 genes were regulated in the presence of E2, native ways of binding to ERα and thus avoid the but the induction of ERβ in these cells resulted in the repulsive interaction. An example of an asymmetric, suppression of almost 1,000 of the ERα-regulated genes68. rigid and non-planar ligand that displays impressive In a similar way to previous observations, ERβ antago- selectivity through the β-face is 8β‑VE2 (REF. 57) (FIG. 2c). nized the ERα-mediated upregulation of proliferative Examples of robust ERβ ligands that achieve significant genes, and in a cell proliferation assay, ERβ completely selectivity through the α-face amino acid difference opposed the ERα-mediated increase in cell number in include the fluroenones developed by Merck58 (FIGS 2d,3c) response to E2 (REF. 68). Overexpression of ERβ in MCF‑7 and the fused chromanols developed by Eli Lilly59 (FIG. 3c). and T47D breast cancer cells resulted in anti-proliferative A final example of a ligand that achieves its selectivity responses in the absence of added ligand69,70. In MCF‑7 through a second shell amino acid difference is the cells, high ERβ expression caused G2 cell cycle arrest in 4‑benzyl chromanol developed by AkzoNobel60 (FIG. 3c). a ligand-independent fashion; ERβ inactivated cyclin- dependent kinase 1 through activation of the tumour Metabolic constraints. A requirement for developing an suppressor genes and kinase inhibitors growth arrest orally active ERβ-selective ligand is reasonable bioavail- and DNA-damage-inducible-α (GADD45A) and BTG2, ability and elimination half-life. Most high-affinity ER and repression of the cyclin B1 gene70. In the MCF‑7 ligands possess a phenolic hydroxyl functional group and T47D human breast cancer cell lines, high ERβ that is prone to rapid phase II glucuronidation that will expression diminished cell proliferation in a ligand- substantially reduce both the oral bioavailability and independent manner by decreasing HER3 (also known half-life. One approach to circumvent this problem is as ERBB3) phosphorylation following heregulin‑β1 acti- to mask the hydroxyl group as a pro-drug with esters or vation of HER2 (also known as ERBB2)–HER3 dimers, ethers that prevent first-pass glucuronidation; the active which subsequently resulted in decreased activation of form of the drug will then be released in vivo through the the PI3K–AKT phosphorylation signalling pathway69. action of esterases or oxidases, respectively61. Furthermore, ERβ upregulated the levels of PTEN, which A second approach is to make it more difficult for the is a tumour suppressor and inhibitor of AKT signalling69. glucuronosyltransferases to catalyse the phase II metabo- In xenograft models of human breast cancer cells, E2 lism. This can be achieved by adding substituents adja- stimulated the development of tumours in the presence cent to the hydroxyl group to sterically hinder the reaction of ERα alone, whereas in the presence of both ERα and and/or by introducing electron withdrawing groups that ERβ, tumour establishment and growth were reduced are orthologous to the hydroxyl group, which will render or prevented64,71,72. Moreover, treatment with the phyto- the hydroxyl group less nucleophilic. Another approach oestrogen liquiritigenin (FIG. 3b), which selectively binds is to use phenol bioisosteres, such as a fused pyrazole62. and activates ERβ, prevented the in vivo establishment and growth of breast tumours, whereas large tumours ERβ: novel target for treatment of cancer developed in animals treated with E2 or the non-selective Breast cancer. ERα is associated with enhanced cell agonist diethylstilbestrol73. proliferation in breast tissue in response to oestrogens, Together, these data support the notion that selective whereas the presence of ERβ exerts anti-proliferative targeting of ERβ with an agonist may be an important effects. Moreover, the presence of ERβ seems to poten- therapeutic strategy for the clinical management of tiate the anti-tumorigenic efficacy of tamoxifen, and breast cancer. ERα-selective antagonists may also hold immunohistochemical analysis of archival breast cancer therapeutic promise, and combination treatments with samples from women treated with adjuvant tamoxifen an ERβ agonist and an ERα antagonist might be an opti- suggested that the presence of ERβ may be associated mal approach for the treatment of hormone-responsive with significantly improved patient survival63–66. breast cancer. In the mouse mammary epithelial cell line HC‑11, which expresses endogenous ERα and ERβ, silencing of Benign prostate hyperplasia and prostate cancer. Changes ERβ expression resulted in an ERα-mediated prolifera- in the testosterone/oestrogen ratio in ageing men have tive response to E2 and allowed the cells to grow in soft been strongly implicated in the development of benign

and malignant prostate disease74. Studies in knockout of mesothelioma cell proliferation includes reduced mice and the use of ER subtype-selective agonists have epidermal growth factor receptor signalling and the identified ERα as the ER subtype responsible for medi- subsequent inhibition of the AKT and MAPK signal- ating inflammatory and pre-malignant pathologies. By ling pathways. However, data on the expression of ERβ contrast, ERβ is associated with anti-proliferative and in mesothelioma tumour samples is based on a small anti-inflammatory activity74,75. In a recent study, xeno- patient study, and the role of ERβ in regulating meso- grafted benign prostate hyperplasia (BPH) and prostate thelioma growth and tumour progression in vivo needs cancer tissues from men who had undergone surgery further investigation. were treated with the ERβ-selective agonist 8β‑VE2 (FIG. 3c). The results showed a significant induction of T-cell lymphoma. ERβ is the predominant ER expressed stromal and epithelial apoptosis, suggesting a therapeutic in white blood cells and in several human lymphoma potential of selective activation of ERβ for the treatment cells84. Treatment of murine and human T‑cell lym- of prostate cancer and BPH76. phoma cell lines with selective activators of ERβ inhib- Analogous to treatment options for breast cancer, an ited their proliferation and induced apoptosis84,85. ERα-selective antagonist alone or in combination with Furthermore, administration of selective ERβ agonists an ERβ agonist could be a promising approach for treat- to male and female mice inoculated with mouse EG7 ment of BPH and prostate cancer. T‑cell lymphoma cells resulted in significant inhibition of lymphoma establishment and growth, hinting at the Colon cancer. The risk for development of colorectal therapeutic potential of selective ERβ agonists for the cancer (CRC) is higher in men than in women, and treatment of lymphoma84. women on MHT have a significantly reduced relative risk of CRC compared to women who do not use MHT. ERα and ERβ in the cardiovascular system Moreover, a further risk reduction is associated with The overall risk for cardiovascular disease is generally increased duration of therapy77,78. lower in premenopausal women than in age-matched ERβ is the predominant ER subtype in colorectal men; premenopausal women have a decreased incidence mucosal epithelium, and loss of ERβ leads to a hyperpro- of left ventricular hypertrophy (LVH), CHD and cardiac liferative state and aberrant differentiation and apoptosis79. remodelling after myocardial infarction. However, cessa- The role and mechanism behind the protective effect of tion in ovarian oestrogen production during menopausal ERβ has been examined in different colon cancer cell transition dramatically increases their risk of experi- lines and experimental systems77,80,81. E2 treatment of ova- encing cardiovascular events, and in postmenopausal riectomized (OVX) mice had protective effects against women the risk for heart disease rapidly approaches the the development of preneoplastic lesions in wild-type risk level observed in men86,87. mice but not in ERβ-knockout mice, suggesting that Both ER subtypes are expressed in vascular endothe- ERβ is an important target in the prevention of tumour lial cells, smooth muscle cells and myocardial cells. The formation in the colon81. Overexpression of ERβ in the EDC ligand40, which is excluded from the nucleus, was SW480 human colon adenocarcinoma cell line resulted used to study carotid artery re-endothelialization in in inhibition of cell proliferation in vitro and in a sig- OVX mice following injury36. Both E2 and EDC resulted nificant reduction of tumour formation in vivo. These in a similar regrowth of the endothelial cell layer after results were explained by the ERβ-mediated suppression injury, suggesting that cytoplasmic ER signalling is suf- of cell-cycle promoting genes and stimulation of path- ficient to preserve the integrity of the vascular endothe- ways that inhibit cell cycle progression80. lial cell layer. Moreover, compared with wild-type mice, In summary, these studies suggest that ERβ mediates ERα-knockout mice showed negligible endothelial pro- MHT protection against colorectal tumour formation and tection against injury, suggesting that ERα is the primary that ERβ-selective agonists could be useful for preventing mediator of E2‑dependent re-endothelialization36. and/or treating CRC. E2 administration has been shown to cause rapid vascular wall dilatation through the stimulation of Mesothelioma. Expression of ERβ in tissue samples nitric oxide production in wild-type female OVX mice from patients with mesothelioma is considered to be a but not in mice lacking ERα or ERβ88. Genetic deletion prognostic factor of better survival82. Established human of ERβ resulted in an age-dependent development of malignant mesothelioma cell lines express varying hypertension in both male and female mice, possibly levels of ERβ but not ERα. Like in breast cancer cells, owing to loss of ERβ-dependent stimulation of nitric ligand activation of ERβ in mesothelioma cells inhib- oxide production combined with abnormalities in vas- ited cell proliferation. Knockdown of ERβ expression cular smooth muscle ion channel function89. Treatment in malignant mesothelioma cells led to a more inva- effects of ER subtype-selective agonists have been stud- sive phenotype, loss of cell–cell contact inhibition, ied in two models of hypertension. In the aldosterone anchorage-independent growth and the formation of plus high salt-treated (AST) OVX rat model, activation colonies in soft agar. Conversely, re-expression of ERβ of ERα with the selective agonist 16α‑LE2 (FIG. 3a) or acti- resulted in a more differentiated, epitheloid-like phe- vation of ERβ with the selective agonist 8β‑VE2 (FIG. 3c) notype, increased anchorage-dependent growth of the resulted in decreased blood pressure compared with cells and down-modulation of cell-cycle-promoting vehicle-treated (control) AST rats90. In the spontaneously genes83. The mechanism of ERβ-mediated inhibition hypertensive rat (SHR) model, however, only E2 and

8β‑VE2 elicited lowering of blood pressure compared to effect, suggesting that some of the protective effects vehicle treated OVX SHRs. In addition, 8β‑VE2 lowered of E2 were ERα independent98. E2 treatment of ath- peripheral vascular resistance91. erogenic diet-fed OVX wild-type and ERα-knockout The cardioprotective role of ERα and ERβ has been mice resulted in a significant attenuation of early ath- investigated following the use of various models and erosclerotic lesion development, irrespective of geno- techniques90,92–96. Induction of experimental myo- type, compared with vehicle-treated OVX controls99. cardial infarction in ERβ-knockout and wild-type Because of the absence of any difference in early ath- OVX mice caused similar infarct size, left ventricular erosclerotic lesion development between wild-type and remodelling and impaired contractile function in both ERα-knockout female mice in response to E2, it was genotypes. However, in contrast to wild-type females, concluded that ERα is dispensable for protection against ERβ-knockout female mice showed increased mortality, the development of early-stage atherosclerotic lesions which was linked to altered Ca2+ homeostasis in their in female mice99. cardiac muscle95. In transverse aorta-constricted wild- Heat shock protein 27 (HSP27; also known as type and ERα- or ERβ-deficient mice, a difference in the HSPβ1) has been suggested to correlate with intima development of pressure overload hypertrophy between thickness100 and to provide protection from coronary males and females was observed96. Wild-type male mice artery disease101. Release of HSP27 has been shown to developed a 64% increase in the heart weight to body be lower from artery segments with atherosclerosis weight (HW/BW) ratio (indicating an increased degree than from healthy artery segments102. Furthermore, of cardiac hypertrophy) compared to non-constricted circulating plasma levels of HSP27 were significantly wild-type male mice, whereas wild-type female mice lower in patients with atherosclerosis than in healthy showed a 31% increase in HW/BW ratio compared subjects, suggesting that HSP27 could serve as a bio- to non-constricted wild-type females96. Moreover, the marker and index of atherosclerosis102. Recently, it increase in HW/BW ratio in female ERβ-knockout was demonstrated that ERβ but not ERα can interact mice was significantly higher than the increase in HW/ specifically with HSP27 and that exposure of human BW ratio in female ERα-knockout mice, suggesting macrophages to E2 or the ERβ-selective agonist DPN that ERβ but not ERα has an important role in attenu- leads to a robust increase in extracellular HSP27 lev- ating cardiac hypertrophy in females. In contrast to els103. Moreover, compared to E2, DPN caused greater ERβ-knockout females, ERβ-knockout male mice release of intracellular HSP27 from macrophages showed no increased degree of cardiac remodelling fol- in vitro following prolonged exposure103. Female mice lowing transverse aortic constriction, implicating a sex overexpressing human HSP27 and deficient for the dimorphic, cardioprotective role of ERβ96. In another, Apoe gene (HSP27o/eApoe–/– mice) developed signifi- similar type of study, male wild-type mice were found cantly less atherosclerotic lesions compared to female to display more signs of heart failure and cardiac fibro- control Apoe–/– mice and male HSP27o/eApoe–/– mice. sis than females in response to pressure overload92. The female to male difference in atherosclerotic lesion This study also reports that the sex differences in the development was explained by the more than tenfold development of LVH and heart failure are related to dif- higher serum levels of HSP27 in female compared to ferent actions of ERβ; ERβ promotes fibrosis in males male mice104. The lesion-protective effect of HSP27 but inhibits fibrosis in females. These male to female is explained by its inhibition of acetylated low den- differences in cardiac remodelling in mice are in agree- sity lipoprotein (acLDL) uptake by macrophages and ment with observations in humans, as the degree of its inhibition of acLDL-mediated release of the pro- cardiac remodelling after myocardial infarction and the inflammatory cytokine interleukin‑1β (IL‑1β). HSP27 HW/BW ratio are significantly higher in men than in also stimulated the release of the anti-inflammatory women92,96. cytokine IL‑10 (REF. 104). Ovariectomy of Apoe–/– and Ovariectomy of wild-type mice caused a significantly HSP27o/eApoe–/– mice resulted in similar degrees of greater degree of cardiac injury following ischaemia– atherosclerotic lesion development, but E2 supple- reperfusion than in intact, cycling, female mice. DPN mentation reversed this effect in both genotypes, with treatment of OVX mice led, however, to significant a more pronounced decrease in lesion development recovery of cardiac functionality compared to vehicle- in the HSP27o/eApoe–/– mice compared to the Apoe–/– treated OVX mice. The recovery caused by DPN mice103. Treatment with the ERβ-selective agonist DPN was explained by ERβ-dependent upregulation of resulted in 28% fewer atherosclerotic lesions in HSP27o/ cardioprotective genes97. eApoe–/– mice than in Apoe–/– mice not overexpressing Epidemiological studies in humans support an HSP27 (REF. 103). The in vivo E2- and DPN-mediated atheroprotective role for E2. To delineate the specific atheroprotective effect was confirmed to involve an ER mediator of E2’s protective effects, various wild-type increased release of HSP27 from macrophages in the and knockout animal models and treatment regimes female HSP27o/eApoe–/– mice103. The role of ERβ in the have been used. In OVX mice deficient in apolipopro- protection against atherosclerotic lesion development tein E (APOE) or both APOE and ERα, ERα was con- was further confirmed by use of the potent ERβ-selective cluded to be the major mediator of the E2 protective agonist 8β‑VE2 (REF. 105). effect in advanced stage atherosclerosis98. However, the In summary, both ERα and ERβ mediate the cardiopro- extent of atheroprotection by ERα exceeded what could tective effects of E2, but probably by separate mechanisms be explained from its lipid- and cholesterol-lowering and in different types of injuries.

ERs in multiple sclerosis and Alzheimer’s disease preservation109,116. Despite these therapeutic effects, the Multiple sclerosis. Multiple sclerosis is an inflammatory, side-effect profile (increased risk of thromboembolic demyelinating and neurodegenerative disease affecting events, breast cancer and endometrial cancer) of ERα- approximately 1 million to 2 million people worldwide selective agonists has prevented their development for and with a prevalence in women-to-men of between the treatment of multiple sclerosis. 2:1 and 3:1 (REFS 106,107). Symptoms of the disease in Treatment of a mouse EAE model with the selective women go in relapse–remittance cycles107,108: there is a ERβ agonist DPN did not have anti-inflammatory activity clinical improvement in symptoms during pregnancy, but promoted recovery during the chronic phase, when oestrogen (primarily E3) levels are high, but reducing demyelination and preserving the number of patients suffer a relapse postpartum, when there is a axons in white matter as well as decreasing neuronal dramatic drop in E3 levels. This correlation implicates abnormalities in grey matter. Moreover, DPN treatment E3 in the prevention of multiple sclerosis relapse107. elicited a significant recovery of motor performance116. Multiple sclerosis has primarily been considered A more detailed study of the effects of selective agonist to be a T‑cell-dependent autoimmune disease that is activation of ERβ revealed that ERβ stimulates intrinsic associated with the production of pro-inflammatory axon repair mechanisms despite the continued pres- cytokines (for example, interferon-γ (IFNγ) and IL‑17) ence of inflammation110. Treatment of EAE mice with and pro-inflammatory chemokines (for example, mono- DPN resulted in neuroprotective effects by increasing cyte chemotactic protein 1 (MCP1; also known as C-C the number of differentiated oligodendrocytes, which motif chemokine 2) and RANTES (also known as C-C motif subsequently also led to improved axon myelination. chemokine 5)) during disease progression, and the Functionally, this in turn led to the restoration of axon production of anti-inflammatory cytokines (for exam- conduction velocity and axon refractoriness110. As DPN ple, IL‑4 and IL‑10) during recovery. In addition to T did not modulate the inflammatory response, the com- cells, B cells have been reported to have a key role in binatory effect of the anti-inflammatory agent IFNβ and multiple sclerosis pathogenesis and in the protec- the ERβ-selective agonist DPN was investigated in the tive effect of E2 (REF. 109). The rationale for current EAE model117. Combination of the two agents resulted multiple sclerosis therapies is focused on dampening in an additive decrease in disease severity with axons inflammation, and although the anti-inflammatory spared at both early and late stages of the disease. In a and immunomodulatory treatments used today are effec- study using an ERβ-selective agonist, chloroindazole118 tive in reducing relapse rates, their disease-modifying (FIG. 3b) was shown to have anti-inflammatory activity, properties (prevention of neurodegeneration, axon loss inhibiting the development of multiple sclerosis in the and clinical disability) are poor. Preservation of axon EAE model31. Whether this ligand also has neuroprotec- integrity is vital for preventing disease progression and tive and remyelinating effects that are similar to those of terminal neurodegeneration. Future therapy develop- DPN was not disclosed. ment is therefore focused on identifying agents that In summary, the combination of an anti-inflammatory can inhibit demyelination and/or stimulate remyelina- agent with a neuroprotective ERβ agonist or the com- tion, thus preventing axon degeneration and clinical bination of two ERβ agonists, one with primarily anti- disability110. inflammatory activity and the other with neuroprotective E3 was shown to be effective in the treatment of and axon remyelinating characteristics — may present an multiple sclerosis symptoms in a pilot clinical trial on important disease modifying strategy for the treatment women with relapsing–remitting multiple sclerosis and of multiple sclerosis. in the EAE mouse model of multiple sclerosis in both males and females106,107,111,112. Studies that used EAE mice Alzheimer’s disease. The incidence of Alzheimer’s disease deficient in ERα or ERβ expression or that used classi- is greater in women than in men, and it is the leading cal SERMs or ER subtype-selective agonists have shown cause of dementia in the world, estimated to afflict 20 that ERα is the most effective ER subtype in the preven- million to 30 million people worldwide. Alzheimer’s tion and treatment of multiple sclerosis and that both disease most often begins with occasional, minor lapses ERα agonists and classical SERMs can suppress disease in recalling recent events of daily life, which is referred symptoms113–116. Selective activation of ERα has curative to as mild cognitive impairment. With time, additional effects at onset of the disease, mirroring all protective deficits become noticeable, including difficulties in per- effects observed with E2 treatment. Agonist-activated forming complex tasks, difficulties in finding words, ERα mediates anti-inflammatory effects by decreasing emotional lability or spatial disorientation. Several years T‑cell-dependent pro-inflammatory cytokine produc- to decades after the initial amnestic symptoms, patients tion and by increasing the number of T‑regulatory cells, with Alzheimer’s disease experience marked dementia Oligodendrocytes Branched glia that are involved which have immune-suppressive effects. In addition, with profound memory impairment, global cognitive 119 in neuronal maintenance and ERα was recently shown to mediate upregulation of deficits and disorientation . support. They produce myelin, the transmembrane protein programmed cell death 1 The pathological hallmarks of Alzheimer’s disease which insulates neuronal ligand 1 (PDL1) on B cells and to increase the number are the assembly and deposition of neurotoxic aggre- processes. of IL‑10‑producing regulatory B cells, which have an gates of amyloid-β (Aβ) into extracellular plaques and Axon refractoriness important protective role during the initial stage of mul- the formation of intraneuronal filaments of hyper- A measure of axon tiple sclerosis. Moreover, ERα also elicits neuroprotective phosphorylated forms of the microtubule-associated responsiveness to stimuli. effects, as evidenced by reduced demyelination and axon protein tau. Moreover, recent evidence suggests that

&GITCFCVKQPQH cannot prevent the initiation or development of the #βRGRVKFGCIITGICVGU #βCIITGICVGU disease, and neither do they have disease-modifying properties (that is, they cannot promote the clearance +&'CPF0'2↑ of Aβ from existing plaques and repair neuronal damage caused by Aβ deposition). Several lines of evidence suggest that age-related loss of sex hormones increases the risk for the development of Alzheimer’s disease in both women and men121. Brain /KETQINKCNEGNNCEVKXKV[↑ and/or plasma levels of E2 have been reported to be lower in women with Alzheimer’s disease as compared ' to age-matched healthy women121,122. However, therapy #β with CEE plus MPA or CEE alone was shown to increase the incidence of dementia in women aged 65 years or older, and therefore it was concluded that MHT can- βUGETGVCUG #22 αUGETGVCUG↑ not be recommended for women of that age group for prevention against Alzheimer’s disease123. However, the WHI study has been heavily criticized for the high aver- age age of women enrolled in the study; two-thirds of

%GNNOGODT the participants were between the ages of 60 and 70. The concern is that women who have lived 10–20 years in γUGETGVCUG γUGETGVCUG a hypo-oestrogenic state have accumulated neuronal damage that cannot be repaired by oestrogen therapy. CPG It has been suggested that MHT should be initiated during menopausal transition or soon after the menopause while neurons are still healthy and able to benefit from oestrogen exposure124. E2 is reported to have neuroprotective effects and to regulate Aβ accumulation through activation of ERα or Figure 4 | Alzheimer’s disease. The membrane-spanning amyloid precursor protein ERβ or both121. E2 protects against Aβ-induced neuro- (APP) is proteolytically cleaved by α‑, β‑ and γ‑secretases. Cleavage by β‑secretase generates the aggregating and plaque-forming amyloid- (A ) peptide, accumulation toxicity by various mechanisms: through the regulation of 0CVWTβ G4Gβ XKGYU^&TWI&KUEQXGT[ of which eventually leads to the development of Alzheimer’s disease. Oestradiol (E2) has genes involved in apoptotic and anti-apoptotic pathways; 2+ been shown to protect against Aβ-induced neurotoxicity by: upregulating the expression through the regulation of intracellular Ca accumulation; of α‑secretase, thereby promoting the non-amyloidogenic cleavage of APP; increasing through anti-inflammatory action; and through an inhib- the expression of the two Aβ-degrading enzymes (insulin-degrading enzyme (IDE) and itory effect on tau hyperphosphorylation. With respect neprilysin (NEP)); and stimulating microglial internalization and phagocytosis of Aβ, to Aβ accumulation, E2 appears to be involved in both thereby promoting clearance of existing aggregated and deposited Aβ peptide. Aβ production and clearance. E2 promotes the non-amyloidogenic cleavage of APP by increasing the expression of α‑secretase, and it promotes Aβ clearance by increas- inflammatory processes are also involved in the patho- ing the internalization and phagocytosis of Aβ by micro- genesis of Alzheimer’s disease119,120. Proteolytic frag- glia, as well as by upregulating insulin-degrading enzyme mentation of membrane-bound amyloid precursor (IDE) and neprilysin (NEP), two enzymes that have protein (APP) is sequentially generated by the activity Aβ-degrading properties121,122,125 (FIG. 4). A 30% decrease of α-secretase, β-secretase and γ-secretase, respectively; in NEP activity was observed in OVX rats, but the activ- it is the β‑secretase that gives rise to the plaque-forming ity was restored by oestrogen replacement. In the human Aβ (FIG. 4). The familial, hereditary early-onset form of neuroblastoma cell line SH‑SY5Y, E2 or the ER subtype- Alzheimer’s disease is caused by dominant missense selective agonists PPT or DPN increased the expression mutations in APP or in presenilin 1 (PSEN1) and PSEN2 of NEP in a dose-dependent manner. The subtype- (which encode γ‑secretase components); these muta- selective agonists were, however, less efficacious than E2 tions increase the tendency of Aβ to aggregate and thus (REF. 126), suggesting that perhaps both ER subtypes result in the formation of Aβ plaques119,120. Additionally, need to be activated to get maximal NEP gene activa- the ε4 allele of the cholesterol transport protein APOE tion. IDE is expressed at high concentrations in the has been associated with greater hippocampal atrophy brain, and besides its role in degrading insulin and and memory impairment in women121. other peptides, it is also involved in the degradation Existing Alzheimer’s disease therapies are largely and clearance of Aβ. In an oestrogen-deficient trans- symptomatic and include acetylcholinesterase inhibitors genic Alzheimer’s disease mouse model (APP23/Ar+/–), (which are aimed to improve memory and cognition), decreased levels of IDE correlated with the early appear- psychotropic drugs (which are aimed to treat behavioural ance of Aβ plaques in these mice compared with APP23 symptoms, for example, apathy, anxiety, aggressiveness transgenic control mice, which have normal plasma and and/or depression), neuroprotective drugs and inhibitors brain E2 levels122. In primary cultures of rat hippocampal of β‑secretase (which are aimed to prevent the formation neurons, as well as in hippocampal areas of the brains of plaque-forming Aβ)119. Currently available therapies of adult wild-type female rats and triple transgenic

Alzheimer’s disease model (3×Tg-AD) female mice, genes and thereby elicit different biological effects, as E2 treatment resulted in a robust induction of IDE demonstrated by the chloroindazole ligand (FIG. 3b) expression127. It was further demonstrated that E2 compared to other ERβ-selective agonists31, suggests activation of IDE expression was mediated through that more ERβ subtype-selective agonists with different ERβ and not ERα. Moreover, E2‑mediated induction substituents and/or of diverse structural classes need to of IDE expression in vivo was specifically located to the be characterized and evaluated in various animal disease hippocampal region in wild-type female OVX rats and models to identify the optimal ERβ agonist for a particular in female mouse models of Alzheimer’s disease. In the symptom or disease. female mouse models of Alzheimer’s disease, an age- One approach to secure optimal efficacy of ERβ- related decrease in IDE expression was accompanied mediated therapy could be the use of a combination of by gradual increase in hippocampal Aβ accumulation synthetic ERβ-selective agonists, each having a unique and plaque formation127. However, administration of E2 gene regulatory profile. The MF101 extract, which is to OVX female mouse models of Alzheimer’s disease at an ERβ-selective mixture of 22 different herbs used in the age of 12 months significantly reduced the Aβ load Chinese medicine, is in clinical development for the and attenuated plaque formation in the hippocampus treatment of postmenopausal vasomotor symptoms128, compared to controls. The E2‑mediated decrease in Aβ and a combination of phyto-oestrogens is being evalu- content and plaque area was accompanied by increased ated for its therapeutic potential to reduce the increased hippocampal IDE immunoreactivity127, suggesting that risk of cognitive decline and neurodegenerative disease ERβ has an important role in the prevention of Aβ (for example, Alzheimer’s disease) associated with the accumulation and plaque formation. menopause129. In summary, available data suggest that selective acti- As far as we know, there are no adverse events asso- vation of ERβ may represent a safe, disease-modifying ciated with selective activation of ERβ, and as ERβ is therapy for the treatment of Alzheimer’s disease, espe- not involved in stimulation of breast or uterine tissues cially if initiated during menopausal transition or soon in response to E2 or ERβ-selective agonists, ERβ is a after the menopause. more attractive and safer target for drug development and the treatment of symptoms and diseases than ERα. Conclusion and future directions Moreover, the lack of uterotrophic activity by selec- The inclusion of women 60 years of age and older in the tive ERβ agonists eliminates the need to combine such WHI study explains, at least in part, the negative out- agonists with a progestagen and the side effects that come of that study with respect to the protective role could result; progestagens could increase the risk for of oestrogens against the development of cardiovascular breast cancer and might inhibit desired ERβ-mediated and neurodegenerative diseases. More clinical studies effects at other sites (for example, in the cardiovas- on women from perimenopause up to less than 10 years cular system and central nervous system). ERβ has, postmenopause are urgently needed. however, not yet become a validated target for the treat- Although most genes are commonly regulated by ment of various symptoms and diseases and will not ERβ-selective agonists, it is becoming evident that dif- become so until an ERβ-selective therapeutic agent ferent ERβ agonists display gene-specific effects31,52 that has been approved for clinical use. However, from a may have therapeutic implications. Further in-depth preclinical perspective, the future for selective ERβ studies of ERβ signalling pathways and mechanisms of agonists looks promising, as these drugs have shown gene regulation will be important for our understand- efficacy in various animal models of cancer, cardiovascu- ing of their biological roles in health and disease. The lar disease, multiple sclerosis and Alzheimer’s disease, as fact that different ERβ-selective agonists regulate distinct well as in other indications not discussed in this Review.

1. Boon, W. C., Chow, J. D. & Simpson, E. R. The multiple 8. Mitterling, K. L. et al. Cellular and subcellular 15. Santen, R. J. et al. Postmenopausal hormone roles of estrogens and the enzyme aromatase. localization of estrogen and progestin receptor therapy: an Endocrine Society scientific Prog. Brain Res. 181, 209–232 (2010). immunoreactivities in the mouse hippocampus. statement. J. Clin. Endocrinol. Metab. 95, 2. Drummond, A. E. & Fuller, P. J. The importance of ERβ J. Comp. Neurol. 518, 2729–2743 (2010). S1–S66 (2010). signalling in the ovary. J. Endocrinol. 205, 15–23 (2010). 9. Waters, E. M. et al. Estrogen and aging affect 16. Leitman, D. C. et al. Regulation of specific target 3. Taylor, A. H. & Al-Azzawi, F. Immunolocalisation the synaptic distribution of estrogen receptor genes and biological responses by estrogen receptor of oestrogen receptor β in human tissues. β-immunoreactivity in the CA1 region of female rat subtype agonists. Curr. Opin. Pharmacol. 10, J. Mol. Endocrinol. 24, 145–155 (2000). hippocampus. Brain Res. 1379, 86–97 (2011). 629–636 (2010). 4. Saji, S. et al. Estrogen receptors α and β in the rodent 10. Rossouw, J. E. et al. Risks and benefits of estrogen plus This study describes the classification of mammary gland. Proc. Natl Acad. Sci. USA 97, progestin in healthy postmenopausal women: principal ERβ-selective agonists based on their potency 337–342 (2000). results from the Women’s Health Initiative randomized and efficacy. It discusses the differences 5. Rody, A. et al. Methylation of estrogen receptor β controlled trial. JAMA 288, 321–333 (2002). between the two ER subtypes with respect promoter correlates with loss of ER‑β expression in 11. Clarkson, T. B. & Appt, S. E. Controversies about to gene regulation and the therapeutic mammary carcinoma and is an early indication marker HRT—lessons from monkey models. Maturitas 51, opportunities offered by selective targeting in premalignant lesions. Endocr. Relat. Cancer 12, 64–74 (2005). of the two ER subtypes. 903–916 (2005). 12. Harman, S. M. et al. Timing and duration of menopausal 17. Frasor, J. et al. Response-specific and 6. Yang, X., Li, Y. Z., Mao, Z., Gu, P. & Shang, M. hormone treatment may affect cardiovascular outcomes. ligand dose-dependent modulation of Effects of estrogen and tibolone on bladder histology Am. J. Med. 124, 199–205 (2011). estrogen receptor (ER) α activity by ERβ in and estrogen receptors in rats. Chin. Med. J. 122, 13. Mendelsohn, M. E. Estrogen actions in the cardiovas the uterus. Endocrinology 144, 3159–3166 381–385 (2009). cular system. Climacteric 12 (Suppl. 1), 18–21 (2009). (2003). 7. Gorres, B. K., Bomhoff, G. L., Gupte, A. A. & Geiger, P. C. 14. Tsai, S. A., Stefanick, M. L. & Stafford, R. S. Trends in This report presents data on the different Altered estrogen receptor expression in skeletal muscle menopausal hormone therapy use of US office-based biological responses in the uterus to an and adipose tissue of female rats fed a high-fat diet. physicians, 2000–2009. Menopause 18, 385–392 ERα-selective agonist and an ERβ-selective J. Appl. Physiol. 110, 1046–1053 (2011). (2011). agonist.

18. Helguero, L. A., Faulds, M. H., Gustafsson, J.-Å. 36. Chambliss, K. L. et al. Non-nuclear estrogen receptor 59. Richardson, T. I. et al. Benzopyrans as selective & Haldosen, L. A. Estrogen receptors alfa (ERα) α signaling promotes cardiovascular protection estrogen receptor β agonists (SERBAs). Part 5: and β (ERβ) differentially regulate proliferation and but not uterine or breast cancer growth in mice. combined A‑ and C‑ring structure-activity relationship apoptosis of the normal murine mammary epithelial J. Clin. Invest. 120, 2319–2330 (2010). studies. Bioorg. Med. Chem. Lett. 17, 5563–5566 cell line HC11. Oncogene 24, 6605–6616 (2005). 37. Wu, Q., Chambliss, K., Umetani, M., Mineo, C. & (2007). This study presents data on the opposing effects of Shaul, P. W. Non-nuclear estrogen receptor signaling 60. Veeneman, G. H. & Teerhuis, N. M. Chroman ERα and ERβ on proliferation and apoptosis in the in endothelium. J. Biol. Chem. 286, 14737–14743 derivatives as estrogenic compounds. US patent murine mammary cell line HC11. (2011). 0069303 (2003). 19. Gronemeyer, H., Gustafsson, J.-Å. & Laudet, V. 38. Le Romancer, M., Treilleux, I., Bouchekioua- 61. Ferriz, J. M. & Vinsova, J. Prodrug design of phenolic Principles for modulation of the nuclear receptor Bouzaghou, K., Sentis, S. & Corbo, L. Methylation, drugs. Curr. Pharm. Des. 16, 2033–2052 (2010). superfamily. Nature Rev. Drug Discov. 3, 950–964 a key step for nongenomic estrogen signaling in 62. Wilkening, R. R. et al. Estrogen receptor β-subtype (2004). breast tumors. Steroids 75, 560–564 (2010). selective tetrahydrofluorenones: use of a fused 20. Nilsson, S. & Gustafsson, J.‑Å. in Nuclear Receptors: 39. Madak-Erdogan, Z. et al. Nuclear and extranuclear pyrazole as a phenol bioisostere. Bioorg. Med. Current Concepts and Future Challenges (Proteins and pathway inputs in the regulation of global gene Chem. Lett. 16, 3896–3901 (2006). Cell Regulation) (eds Bunce, C. M. & Campbell, M. J.) expression by estrogen receptors. Mol. Endocrinol. 63. Gruvberger-Saal, S. K. et al. Estrogen receptor β 91–141 (Springer, 2010). 22, 2116–2127 (2008). expression is associated with tamoxifen response in 21. Powell, E., Wang, Y., Shapiro, D. J. & Xu, W. 40. Harrington, W. R. et al. Estrogen dendrimer ERα-negative breast carcinoma. Clin. Cancer Res. Differential requirements of Hsp90 and DNA for the conjugates that preferentially activate extranuclear, 13, 1987–1994 (2007). formation of estrogen receptor homodimers and nongenomic versus genomic pathways of estrogen 64. Hartman, J., Strom, A. & Gustafsson, J.‑Å. Estrogen heterodimers. J. Biol. Chem. 285, 16125–16134 (2010). action. Mol. Endocrinol. 20, 491–502 (2006). receptor β in breast cancer—diagnostic and therapeutic 22. Safe, S. & Kim, K. Non-classical genomic estrogen 41. Amazit, L. et al. Regulation of SRC‑3 implications. Steroids 74, 635–641 (2009). receptor (ER)/specificity protein and ER/activating intercompartmental dynamics by estrogen receptor 65. Honma, N. et al. Clinical importance of estrogen protein‑1 signaling pathways. J. Mol. Endocrinol. 41, and phosphorylation. Mol. Cell. Biol. 27, 6913–6932 receptor-β evaluation in breast cancer patients treated 263–275 (2008). (2007). with adjuvant tamoxifen therapy. J. Clin. Oncol. 26, 23. Hall, J. M. & McDonnell, D. P. Coregulators in nuclear 42. Turjanski, A. G., Vaque, J. P. & Gutkind, J. S. 3727–3734 (2008). estrogen receptor action: from concept to therapeutic MAP kinases and the control of nuclear events. 66. Wu, X. et al. Estrogen receptor-β sensitizes breast targeting. Mol. Interv. 5, 343–357 (2005). Oncogene 26, 3240–3253 (2007). cancer cells to the anti-estrogenic actions of A review of co-activator and co-repressor 43. Zhang, L. et al. Estrogen receptor β-selective agonists endoxifen. Breast Cancer Res. 13, R27 (2011). interactions with ER and their importance for the stimulate calcium oscillations in human and mouse 67. Chang, E. C. et al. Estrogen receptors α and β as biological responses to agonists, antagonists and embryonic stem cell-derived neurons. PLoS ONE 5, determinants of gene expression: influence of ligand, SERMs. e11791 (2010). dose, and chromatin binding. Mol. Endocrinol. 22, 24. Shang, Y., Hu, X., DiRenzo, J., Lazar, M. A. & Brown, M. 44. Zhao, L. & Brinton, R. D. Estrogen receptor α and β 1032–1043 (2008). Cofactor dynamics and sufficiency in estrogen receptor- differentially regulate intracellular Ca2+ dynamics 68. Williams, C., Edvardsson, K., Lewandowski, S. A., regulated transcription. Cell 103, 843–852 (2000). leading to ERK phosphorylation and estrogen Strom, A. & Gustafsson, J.‑Å. A genome-wide study 25. Powell, E. & Xu, W. Intermolecular interactions identify neuroprotection in hippocampal neurons. Brain Res. of the repressive effects of estrogen receptor β on ligand-selective activity of estrogen receptor α/β 1172, 48–59 (2007). estrogen receptor α signaling in breast cancer cells. dimers. Proc. Natl Acad. Sci. USA 105, 45. Anstead, G. M., Carlson, K. E. & Katzenellenbogen, Oncogene 27, 1019–1032 (2008). 19012–19017 (2008). J. A. The estradiol pharmacophore: ligand structure- 69. Lindberg, K., Helguero, L. A., Omoto, Y., Gustafsson, 26. Monroe, D. G. et al. Estrogen receptor α and β estrogen receptor binding affinity relationships and J.‑Å. & Haldosen, L. A. Estrogen receptor β represses heterodimers exert unique effects on estrogen- and a model for the receptor binding site. Steroids 62, Akt signaling in breast cancer cells via downregulation tamoxifen-dependent gene expression in human 268–303 (1997). of HER2/HER3 and upregulation of PTEN — U2OS osteosarcoma cells. Mol. Endocrinol. 19, 46. Brzozowski, A. M. et al. Molecular basis of agonism implications for tamoxifen sensitivity. Breast Cancer 1555–1568 (2005). and antagonism in the oestrogen receptor. Nature Res. 13, R43 (2011). This study demonstrates that ERα and ERβ 389, 753–758 (1997). 70. Paruthiyil, S. et al. Estrogen receptor β causes a G2 heterodimers regulate a unique set of genes, which A landmark paper that provided the foundation for cell cycle arrest by inhibiting CDK1 activity through are distinct from the genes regulated by ERα and the subsequent structure-based design of ER ligands. the regulation of cyclin B1, GADD45A, and BTG2. ERβ homodimers, in a ligand-dependent fashion. 47. Minutolo, F., Macchia, M., Katzenellenbogen, B. S. & Breast Cancer Res. Treat. 1 Dec 2010 (doi:10.1007/ 27. Charn, T. H. et al. Genome-wide dynamics of Katzenellenbogen, J. A. Estrogen receptor β ligands: s10549‑010‑1273‑5). chromatin binding of estrogen receptors α and β: recent advances and biomedical applications. 71. Cvoro, A. et al. Selective activation of estrogen mutual restriction and competitive site selection. Med. Res. Rev. 31, 364–442 (2011). receptor-β transcriptional pathways by an herbal Mol. Endocrinol. 24, 47–59 (2010). A thorough review of the known chemical classes extract. Endocrinology 148, 538–547 (2007). 28. Perissi, V. & Rosenfeld, M. G. Controlling nuclear of ERβ-selective ligands, their pharmacophores 72. Hartman, J. et al. Estrogen receptor β inhibits receptors: the circular logic of cofactor cycles. and their potential therapeutic applications. angiogenesis and growth of T47D breast cancer Nature Rev. Mol. Cell Biol. 6, 542–554 (2005). 48. Mohler, M. L. et al. Estrogen receptor β selective xenografts. Cancer Res. 66, 11207–11213 (2006). 29. Bramlett, K. S., Wu, Y. & Burris, T. P. Ligands specify nonsteroidal estrogens: seeking clinical indications. 73. Mersereau, J. E. et al. Liquiritigenin is a plant-derived coactivator nuclear receptor (NR) box affinity for Expert Opin. Ther. Pat. 20, 507–534 (2010). highly selective estrogen receptor β agonist. Mol. Cell. estrogen receptor subtypes. Mol. Endocrinol. 15, 49. Kuiper, G. G. et al. Comparison of the ligand binding Endocrinol. 283, 49–57 (2008). 909–922 (2001). specificity and transcript tissue distribution of 74. Ellem, S. J. & Risbridger, G. P. Aromatase and 30. Gee, A. C., Carlson, K. E., Martini, P. G., estrogen receptors α and β. Endocrinology 138, regulating the estrogen:androgen ratio in the prostate Katzenellenbogen, B. S. & Katzenellenbogen, J. A. 863–870 (1997). gland. J. Steroid Biochem. Mol. Biol. 118, 246–251 Coactivator peptides have a differential stabilizing 50. Meyers, M. J. et al. Estrogen receptor-β potency- (2010). effect on the binding of estrogens and antiestrogens selective ligands: structure-activity relationship studies 75. McPherson, S. J. et al. Essential role for estrogen with the estrogen receptor. Mol. Endocrinol. 13, of diarylpropionitriles and their acetylene and polar receptor β in stromal-epithelial regulation of prostatic 1912–1923 (1999). analogues. J. Med. Chem. 44, 4230–4251 (2001). hyperplasia. Endocrinology 148, 566–574 (2007). 31. Saijo, K., Collier, J. G., Li, A. C., Katzenellenbogen, 51. Bruning, J. B. et al. Coupling of receptor conformation 76. McPherson, S. J. et al. Estrogen receptor-β activated J. A. & Glass, C. K. An ADIOL‑ERβ‑CtBP transrepression and ligand orientation determine graded activity. apoptosis in benign hyperplasia and cancer of the pathway negatively regulates microglia-mediated Nature Chem. Biol. 6, 837–843 (2010). prostate is androgen independent and TNFα inflammation. Cell 145, 584–595 (2011). 52. Paruthiyil, S. et al. Drug and cell type-specific regulation mediated. Proc. Natl Acad. Sci. USA 107, Elucidation of an ERβ-mediated transrepressive of genes with different classes of estrogen receptor 3123–3128 (2010). anti-inflammatory pathway. Evidence is provided β-selective agonists. PLoS ONE 4, e6271 (2009). 77. Kennelly, R., Kavanagh, D. O., Hogan, A. M. & that 5‑androstene‑3β,17β‑diol is a physiologically 53. Gellman, S. H. On the role of methionine residues in Winter, D. C. Oestrogen and the colon: potential relevant ERβ ligand that triggers ERβ-mediated the sequence-independent recognition of nonpolar mechanisms for cancer prevention. Lancet Oncol. anti-inflammatory activity by transrepression, protein surfaces. Biochemistry 30, 6633–6636 (1991). 9, 385–391 (2008). whereas the endogenous ER hormone E2 is inactive 54. Manas, E. S., Xu, Z. B., Unwalla, R. J. & Somers, W. S. 78. Long, M. D., Martin, C. F., Galanko, J. A. & through this pathway. Understanding the selectivity of genistein for human Sandler, R. S. Hormone replacement therapy, oral 32. Arpino, G., Wiechmann, L., Osborne, C. K. & Schiff, R. estrogen receptor-β using X‑ray crystallography and contraceptive use, and distal large bowel cancer: Crosstalk between the estrogen receptor and the HER computational methods. Structure 12, 2197–2207 a population-based case-control study. tyrosine kinase receptor family: molecular mechanism (2004). Am. J. Gastroenterol. 105, 1843–1850 (2010). and clinical implications for endocrine therapy 55. Pike, A. C. et al. Structure of the ligand-binding 79. Wada-Hiraike, O. et al. Role of estrogen receptor β in resistance. Endocr. Rev. 29, 217–233 (2008). domain of oestrogen receptor β in the presence of a colonic epithelium. Proc. Natl Acad. Sci. USA 103, 33. Hammes, S. R. & Levin, E. R. Extranuclear steroid partial agonist and a full antagonist. EMBO J. 18, 2959–2964 (2006). receptors: nature and actions. Endocr. Rev. 28, 4608–4618 (1999). 80. Hartman, J. et al. Tumor repressive functions of 726–741 (2007). 56. Yu, H., Noskov, S. Y. & Roux, B. Two mechanisms of estrogen receptor β in SW480 colon cancer cells. 34. Acconcia, F., Ascenzi, P., Fabozzi, G., Visca, P. & ion selectivity in protein binding sites. Proc. Natl Cancer Res. 69, 6100–6106 (2009). Marino, M. S‑palmitoylation modulates human Acad. Sci. USA 107, 20329–20334 (2010). 81. Weige, C. C., Allred, K. F. & Allred, C. D. Estradiol estrogen receptor-α functions. Biochem. Biophys. Res. 57. Hegele-Hartung, C. et al. Impact of isotype-selective alters cell growth in nonmalignant colonocytes and Commun. 316, 878–883 (2004). estrogen receptor agonists on ovarian function. reduces the formation of preneoplastic lesions in 35. Lu, Q. et al. Striatin assembles a membrane signaling Proc. Natl Acad. Sci. USA 101, 5129–5134 (2004). the colon. Cancer Res. 69, 9118–9124 (2009). complex necessary for rapid, nongenomic activation 58. Blizzard, T. A. Selective estrogen receptor modulator 82. Pinton, G. et al. Estrogen receptor-β affects the of endothelial NO synthase by estrogen receptor α. medicinal chemistry at Merck. A review. Curr. Top. prognosis of human malignant mesothelioma. Proc. Natl Acad. Sci. USA 101, 17126–17131 (2004). Med. Chem. 8, 792–812 (2008). Cancer Res. 69, 4598–4604 (2009).

83. Pinton, G. et al. Estrogen receptor β exerts tumor 103. Rayner, K. et al. Heat shock protein 27 protects 123. Shumaker, S. A. et al. Conjugated equine estrogens repressive functions in human malignant pleural against atherogenesis via an estrogen-dependent and incidence of probable dementia and mild mesothelioma via EGFR inactivation and affects mechanism: role of selective estrogen receptor β cognitive impairment in postmenopausal women: response to gefitinib. PLoS ONE 5, e14110 modulation. Arterioscler. Thromb. Vasc. Biol. 29, Women’s Health Initiative Memory Study. JAMA 291, (2010). 1751–1756 (2009). 2947–2958 (2004). 84. Yakimchuk, K. et al. Effect of ligand-activated estrogen 104. Rayner, K. et al. Extracellular release of the 124. Brinton, R. D. Investigative models for determining receptor β on lymphoma growth in vitro and in vivo. atheroprotective heat shock protein 27 is mediated by hormone therapy-induced outcomes in brain: Leukemia 25, 1103–1110 (2011). estrogen and competitively inhibits acLDL binding to evidence in support of a healthy cell bias of estrogen This study reports that activation of ERβ with scavenger receptor‑A. Circ. Res. 103, 133–141 (2008). action. Ann. N. Y. Acad. Sci. 1052, 57–74 (2005). two structurally distinct ERβ-selective agonists 105. Sun, J. et al. Attenuation of atherogenesis via the anti- 125. Leissring, M. A. et al. Enhanced proteolysis of suppresses murine T‑cell lymphoma tumour growth inflammatory effects of the selective estrogen receptor β-amyloid in APP transgenic mice prevents plaque in vivo by inhibition of proliferation and stimulation β modulator 8β‑VE2. J. Cardiovasc. Pharmacol. formation, secondary pathology, and premature of programmed cell death. 18 Jun 2011 (doi:10.1097/FJC.0b013e318226bd16). death. Neuron 40, 1087–1093 (2003). 85. Yamasaki, M. et al. Genistein induced apoptotic cell 106. Gold, S. M. & Voskuhl, R. R. Estrogen and 126. Liang, K. et al. Estrogen stimulates degradation death in adult T‑cell leukemia cells through estrogen testosterone therapies in multiple sclerosis. of β-amyloid peptide by up-regulating neprilysin. receptors. Biosci. Biotechnol. Biochem. 74, Prog. Brain Res. 175, 239–251 (2009). J. Biol. Chem. 285, 935–942 (2010). 2113–2115 (2010). 107. Soldan, S. S., Alvarez Retuerto, A. I., Sicotte, N. L. 127. Zhao, L. et al. 17β‑Estradiol regulates insulin- 86. Arias-Loza, P. A., Jazbutyte, V. & Pelzer, T. Genetic and & Voskuhl, R. R. Immune modulation in multiple degrading enzyme expression via an ERβ/PI3‑K pharmacologic strategies to determine the function sclerosis patients treated with the pregnancy hormone pathway in hippocampus: relevance to Alzheimer’s of estrogen receptor α and estrogen receptor β in estriol. J. Immunol. 171, 6267–6274 (2003). prevention. Neurobiol. Aging 6 Jan 2010 cardiovascular system. Gend. Med. 5 (Suppl. A), 108. Compston, A. & Coles, A. Multiple sclerosis. Lancet (doi:10.1016/j.neurobiolaging.2009.12.010). 34–45 (2008). 359, 1221–1231 (2002). 128. Grady, D. et al. MF101, a selective estrogen receptor 87. Vitale, C., Mendelsohn, M. E. & Rosano, G. M. 109. Bodhankar, S., Wang, C., Vandenbark, A. A. & Offner, H. β modulator for the treatment of menopausal hot Gender differences in the cardiovascular effect Estrogen-induced protection against experimental flushes: a phase II clinical trial. Menopause 16, of sex hormones. Nature Rev. Cardiol. 6, 532–542 autoimmune encephalomyelitis is abrogated in the 458–465 (2009). (2009). absence of B cells. Eur. J. Immunol. 41, 1165–1175 129. Zhao, L., Mao, Z. & Brinton, R. D. A select 88. Guo, X., Razandi, M., Pedram, A., Kassab, G. & Levin, (2011). combination of clinically relevant phytoestrogens E. R. Estrogen induces vascular wall dilation: 110. Crawford, D. K. et al. Oestrogen receptor β ligand: enhances estrogen receptor β-binding selectivity mediation through kinase signaling to nitric oxide a novel treatment to enhance endogenous functional and neuroprotective activities in vitro and in vivo. and estrogen receptors α and β. J. Biol. Chem. 280, remyelination. Brain 133, 2999–3016 (2010). Endocrinology 150, 770–783 (2009). 19704–19710 (2005). This study shows that treatment of EAE mice 130. Jordan, V. C. Selective estrogen receptor modulation: 89. Zhu, Y. et al. Abnormal vascular function and with the ERβ-selective agonist DPN results in an a personal perspective. Cancer Res. 61, 5683–5687 hypertension in mice deficient in estrogen receptor β. increased number of differentiated oligodendrocytes (2001). Science 295, 505–508 (2002). and a remarkable remyelination and restoration of An excellent review of the SERM concept and the 90. Arias-Loza, P. A. et al. Both estrogen receptor axon conduction velocity and axon refractoriness. clinical applications of SERMs. subtypes, α and β, attenuate cardiovascular 111. Palaszynski, K. M., Liu, H., Loo, K. K. & Voskuhl, R. R. 131. Grigsby, J. G. et al. Effects of tamoxifen versus remodeling in aldosterone salt-treated rats. Estriol treatment ameliorates disease in males raloxifene on retinal capillary endothelial cell Hypertension 50, 432–438 (2007). with experimental autoimmune encephalomyelitis: proliferation. J. Ocul. Pharmacol. Ther. 27, 91. Jazbutyte, V. et al. Ligand-dependent activation of implications for multiple sclerosis. J. Neuroimmunol. 225–233 (2011). ERβ lowers blood pressure and attenuates cardiac 149, 84–89 (2004). 132. Albertazzi, P. & Sharma, S. Urogenital effects of hypertrophy in ovariectomized spontaneously 112. Sicotte, N. L. et al. Treatment of multiple sclerosis with selective estrogen receptor modulators: a systematic hypertensive rats. Cardiovasc. Res. 77, 774–781 the pregnancy hormone estriol. Ann. Neurol. 52, review. Climacteric 8, 214–220 (2005). (2008). 421–428 (2002). 133. Hendrix, S. L. & McNeeley, S. G. Effect of selective 92. Fliegner, D. et al. Female sex and estrogen 113. Elloso, M. M., Phiel, K., Henderson, R. A., estrogen receptor modulators on reproductive tissues receptor-β attenuate cardiac remodeling and Harris, H. A. & Adelman, S. J. Suppression of other than endometrium. Ann. N. Y. Acad. Sci. 949, apoptosis in pressure overload. Am. J. Physiol. experimental autoimmune encephalomyelitis using 243–250 (2001). Regul. Integr. Comp. Physiol. 298, R1597–R1606 estrogen receptor-selective ligands. J. Endocrinol. 134. McDonnell, D. P., Clemm, D. L., Hermann, T., (2010). 185, 243–252 (2005). Goldman, M. E. & Pike, J. W. Analysis of estrogen 93. Kararigas, G., Fliegner, D., Gustafsson, J.‑Å. & 114. Polanczyk, M. et al. Estrogen receptor‑1 (Esr1) and ‑2 receptor function in vitro reveals three distinct classes Regitz-Zagrosek, V. The role of the estrogen/ (Esr2) regulate the severity of clinical experimental of antiestrogens. Mol. Endocrinol. 9, 659–669 (1995). estrogen‑receptor‑β axis in the genomic response allergic encephalomyelitis in male mice. Am. J. Pathol. 135. Shang, Y. & Brown, M. Molecular determinants for the to pressure overload-induced hypertrophy. Physiol. 164, 1915–1924 (2004). tissue specificity of SERMs. Science 295, 2465–2468 Genomics 43, 438–446 (2011). 115. Polanczyk, M. et al. The protective effect of (2002). 94. Pedram, A., Razandi, M., O’Mahony, F., Lubahn, D. 17β‑estradiol on experimental autoimmune 136. Tzukerman, M. T. et al. Human estrogen receptor & Levin, E. R. Estrogen receptor-β prevents cardiac encephalomyelitis is mediated through estrogen transactivational capacity is determined by both fibrosis. Mol. Endocrinol. 24, 2152–2165 (2010). receptor-α. Am. J. Pathol. 163, 1599–1605 (2003). cellular and promoter context and mediated by 95. Pelzer, T. et al. Increased mortality and aggravation of 116. Tiwari-Woodruff, S., Morales, L. B., Lee, R. & two functionally distinct intramolecular regions. heart failure in estrogen receptor-β knockout mice Voskuhl, R. R. Differential neuroprotective and Mol. Endocrinol. 8, 21–30 (1994). after myocardial infarction. Circulation 111, antiinflammatory effects of estrogen receptor (ER)α 137. Shelly, W., Draper, M. W., Krishnan, V., Wong, M. & 1492–1498 (2005). and ERβ ligand treatment. Proc. Natl Acad. Sci. USA Jaffe, R. B. Selective estrogen receptor modulators: 96. Skavdahl, M. et al. Estrogen receptor-β mediates 104, 14813–14818 (2007). an update on recent clinical findings. Obstet. Gynecol. male-female differences in the development of 117. Du, S., Sandoval, F., Trinh, P. & Voskuhl, R. R. Surv. 63, 163–181 (2008). pressure overload hypertrophy. Am. J. Physiol. Additive effects of combination treatment with 138. Barkhem, T. et al. Differential response of estrogen Heart Circ. Physiol. 288, H469–H476 (2005). anti-inflammatory and neuroprotective agents in receptor α and estrogen receptor β to partial estrogen 97. Nikolic, I. et al. Treatment with an estrogen experimental autoimmune encephalomyelitis. agonists/antagonists. Mol. Pharmacol. 54, 105–112 receptor‑β‑selective agonist is cardioprotective. J. Neuroimmunol. 219, 64–74 (2010). (1998). J. Mol. Cell. Cardiol. 42, 769–780 (2007). 118. De Angelis, M., Stossi, F., Carlson, K. A., 139. Schechter, I. & Berger, A. On the size of the active 98. Hodgin, J. B. et al. Estrogen receptor α is a major Katzenellenbogen, B. S. & Katzenellenbogen, J. A. site in proteases. I. Papain. Biochem. Biophys. Res. mediator of 17β‑estradiol’s atheroprotective effects Indazole estrogens: highly selective ligands for the Commun. 27, 157–162 (1967). on lesion size in Apoe–/– mice. J. Clin. Invest. 107, estrogen receptor . J. Med. Chem. 48, 1132–1144 β Acknowledgements 333–340 (2001). (2005). Research in J.‑Å.G’s laboratory is supported by The Swedish 99. Villablanca, A. C. et al. 17β‑estradiol prevents early- 119. Selkoe, D. J. & Schenk, D. Alzheimer’s disease: Cancer Fund, the Welch Foundation (Houston, Texas, USA) and stage atherosclerosis in estrogen receptor-α deficient molecular understanding predicts amyloid-based the Texas Emerging Technology Fund (Austin, Texas, USA) female mice. J. Cardiovasc. Transl. Res. 2, 289–299 therapeutics. Annu. Rev. Pharmacol. Toxicol. 43, under agreement number 300‑9‑1958. (2009). 545–584 (2003). 100. Pockley, A. G., Georgiades, A., Thulin, T., de Faire, U. 120. Vegeto, E., Benedusi, V. & Maggi, A. Estrogen Competing interests statement & Frostegard, J. Serum heat shock protein 70 levels anti-inflammatory activity in brain: a therapeutic The authors declare competing financial interests: see web predict the development of atherosclerosis in subjects opportunity for menopause and neurodegenerative version for details. with established hypertension. Hypertension 42, diseases. Front. Neuroendocrinol. 29, 507–519 235–238 (2003). (2008). FURTHER INFORMATION 101. Zhu, J. et al. Increased serum levels of heat shock 121. Pike, C. J., Carroll, J. C., Rosario, E. R. & Barron, Stefan Nilsson and Konrad Koehler’s homepage: protein 70 are associated with low risk of coronary A. M. Protective actions of sex steroid hormones in http://www.karobio.se artery disease. Arterioscler. Thromb. Vasc. Biol. Alzheimer’s disease. Front. Neuroendocrinol. 30, Jan-Åke Gustafsson’s homepage: http://cnrcs.uh.edu 23, 1055–1059 (2003). 239–258 (2009). 102. Martin-Ventura, J. L. et al. Identification by a 122. Yue, X. et al. Brain estrogen deficiency accelerates Aβ SUPPLEMENTARY INFORMATION differential proteomic approach of heat shock plaque formation in an Alzheimer’s disease animal See online article: S1 (table) | S2 (table) | S3 (figure) protein 27 as a potential marker of atherosclerosis. model. Proc. Natl Acad. Sci. USA 102, 19198–19203 ALL LINKS ARE ACTIVE IN THE ONLINE PDF Circulation 110, 2216–2219 (2004). (2005).