The structural biology of oestrogen metabolism

Highlights ► The structures of oestrogen-metabolising enzymes are discussed and compared. ► Substrate recognition and reaction mechanisms of these enzymes are discussed. ► The impact of structure on inhibitor design is briefly mentioned.

However, these same oestrogen receptors have also been shown (a) to bind to other transcription factors thus influencing the expression of genes that do not contain oestrogen-responsive elements in their promoters, and (b) to engage signal transduction pathways (that may include, but are not limited to, the activation of protein kinases), thus modulating cellular responses to oestrogen. Signal transduction pathways can also be activated by oestrogen binding to cell surface membrane bound receptors [16][17][18].
The three most common estrogens are estrone (E1), estradiol (E2) and estriol (E3). A fourth oestrogen is estetrol (E4). Estradiol is the most potent. Estrone and estradiol are synthesised by the aromatisation of androstenedione and testosterone, respectively ( Fig. 1). They can also be interconverted by the action of 17␤hydroxysteroid dehydrogenases (17␤-HSDs). Estriol is synthesised from estrone via a 16␣-hydroxyestrone intermediate. Although, in some tissues, estrogens can be made on demand, oestrogen can be stored in the form of estrone sulfate. This is synthesised from estrone by the action of oestrogen sulfotransferase with estrone being regenerated by the steroid sulfatase-catalysed hydrolysis of estrone sulfate. A review of many aspects of human steroidogenesis is available [19]. Estrogens are eliminated from the body mainly as the sulfated and glucuronidated derivatives [20, and references therein]. The first step in synthesising these conjugates is the generation of the hydroxylated derivatives. Hydroxylation occurs primarily at the 2-, 4-and 16-positions. The hydroxyl group can then be sulfated, glucuronidated or methylated.
Because there is little published work on estetrol it is discussed only briefly here. It is synthesised in the foetal liver, but its function is presently unknown. Several possible biosynthetic routes have been proposed starting from various androgens, estrogens, and their 3-sulfated derivatives, with the hydroxylations occurring in various different orders and, in the case of the androgens, the aromatisation occurring either before or after the hydroxylations. Evidence suggests that estetrol is made through multiple biosynthetic routes. See [21][22][23][24][25][26][27][28][29] for details. Estetrol will not be discussed further herein.
Given the role of estrogens in preventing, causing and exacerbating disease a good knowledge of how estrogens are synthesised and metabolised may help in the understanding and treatment of disease. This review looks in turn at each of the enzymes involved in oestrogen metabolism in terms of their structure and reaction mechanism, before a final section compares the various enzymes. Inhibition of the enzymes of oestrogen metabolism impacts on the amount of oestrogen in the body and this has been the subject of There are tissue-specific variations in synthetic pathways: see [19]. The A-ring hydroxylated compounds are converted to the methoxy compounds by catechol O-methyltransferases. The structures of estetrol and ethinylestradiol are shown, as is the steroid ring labelling and atom numbering system. multiple therapeutic approaches leading to clinically active agents [30,31]. Also, synthetic estrogens, e.g., ethinylestradiol ( Fig. 1), have been widely employed in hormone replacement therapy [32] and, combined with a progestin, contraception [33,34]. Although inhibitors of these enzymes and synthetic estrogens are not the focus of this review they are briefly discussed, mainly through references to recent reviews. Unless otherwise stated, all discussions below refer to the human enzymes.

Methods
All dockings of ligands into binding sites were carried out using the GOLD software running in Gold Suite v5.0.1. All other manipulations of molecules, including building ligands and preparing proteins for docking runs, were carried out using the Schrödinger software running under Maestro 9.2.109. The protein crystal structures were downloaded from the PDB and run through the Protein Preparation Wizard. Hydrogens were added assuming a neutral pH and the resulting structures were put through a brief minimisation procedure. Visual inspection of these structures was used to identify those residues that form substrate binding sites. All figures of proteins and ligands were prepared using PyMOL (Delano Scientific). Figures of reaction schemes and molecules were prepared using ChemBioDraw Ultra 12.0. Multalin was used to perform sequence alignments (http://multalin.toulouse.inra.fr/multalin/multalin.html).

Oestrogen sulfotransferase
Oestrogen sulfotransferase (EC 2.8.2.4, SULT1E1) is a cytosolic enzyme that catalyzes the sulfonation of estrogens, utilising 3 -phosphoadenosine-5 -phosphosulfate (PAPS) as the sulfatedonating cofactor [35,36] (Fig. 2). This enzyme has a high affinity for estrogens with maximal activity at about K m = 20 nM [37]. There are two other sulfotransferases in the SULT1 class that catalyse the sulfonation of estrogens: SULT1A1 and SULT1A3 (both EC 2.8.2.1, phenol sulfotransferase). However, these enzymes have much lower affinities for estrogens, with maximal activity at about K m = 25 M. For all three human proteins the amino acid sequence is available, as is that of murine SULT1E1 (Fig. 3): SULT1A1 and SULT1A3 are 92.5% identical, and both are approximately 50% identical to SULT1E1. The sequence of murine SULT1E1 is 75% identical to that of the human protein.
Thirty-one crystal structures of twelve different human sulfotransferases have been the subject of a recent review [48] that covers substrate specificity and mechanisms of action. An older review covering sulfotransferase structure and inhibitors is available [49]. Fig. 5 shows the structure of the binding site of human SULT1E1. To generate Fig. 5A the reaction product was docked into the 1HY3 structure after the sulfate had been removed from the cofactor. To generate Fig. 5B the substrate was docked into the 1HY3 structure. In both instances, subsequent to docking, the protein ligand complex was put through 1000 rounds of minimisation prior to preparing the figures. A comparison of the two structures ( Fig. 5C) shows that three of the water molecules have different positions but the ligand moves only slightly, and the surrounding residues barely move at all. The substrate and cofactor binding site takes the form of a tube, open at both ends, running through the core of the protein.
The residues forming the substrate binding site are predominantly hydrophobic and enclose the substrate quite closely. These residues (Y20, F23, P46, F75, F80, C83, K85, M89, K105, H107, F138, F141, V145, A146, G147, H148, Y168, Y239, L242, I246 and M247) are mainly from the three flexible loops (Fig. 4), loop 1 (between helices 5 and 6), loop 2 (between helices 8 and 9), and loop 3 (between helices 15 and 16), but also helices H1 and H8 and sheet Sd. A particularly close interaction is that of the hydrogens attached to C6 and C7 in the B-ring with the face of the Y20 ring. The side chain of F141 points directly at the A-ring of the substrate to have a face-edge interaction. K105 has a hydrogen bond to the oxygen at the 3-position. These interactions position the substrate hydroxyl so that it can act as the acceptor in the sulfate transfer.
A comparison of all human sulfotransferase structures suggests that the binding of substrate and PAPS can occur in a random order, and that substrate binding may be cooperative with prior cofactor binding, but not vice versa [50]. It seems likely that loop 3 must move to enable both substrate and cofactor to bind. The ends of this loop (helices 15 and 16) are in the vicinity of the cofactor: if the entire loop opens as if the ends of the loop act as a hinge then the prior binding of the cofactor may position the remainder of the loop to more easily bind the substrate. This is supported by crystal structures of SULT1A1 in complex with PAP and substrate where the binding site is very similar to that observed when just PAP is bound [47]. (The conformation of this loop subsequent to PAPS binding is also thought to influence substrate binding in human SULT2A1 [51].) There are some reports that at high substrate concentrations sulfotransferases, particularly SULT1A1, suffer from substrate inhibition [46,50]. Whether this is of relevance in vivo is undetermined.
Three conserved residues, K47, H107 and S137, are thought to be involved in the reaction mechanism ( Fig. 6) [38,52]. S137 forms a hydrogen bond to K47, thus preventing the side chain nitrogen of K47 from interacting with the bridging oxygen. This prevents PAPS hydrolysis. Upon substrate binding, the catalytic base H107 abstracts a proton from the substrate hydroxyl, thus enabling nucleophilic attack at the sulphur atom in PAPS. The consequent accumulation of partial negative charge on the bridging oxygen of PAPS disrupts the interaction between S137 and K47, thus enabling K47 to interact with the 5 -phosphate. This assists in the dissociation of the sulfyryl group and completes the catalytic reaction cycle [52,53]. The sulfonation takes place via an in-line displacement mechanism though whether the mechanism is associative (S N 2-like) or dissociative (S N 1-like) is not yet clear [53,54].

Steroid sulfatase
Steroid sulfatase (EC 3.1.6.2, aryl sulfatase C, steryl-sulfatase) catalyzes the conversion of estrone sulfate to estrone (Fig. 1), and dehydroepiandrosterone sulfate (DHEAS) to dehydroepiandrosterone (DHEA) (the reverse of the reaction shown in Fig. 2). The protein is found in the lumen of the endoplasmic reticulum and, in terms of tissue distribution, is probably ubiquitous at low levels, with greater amounts in those tissues associated with reproduction. The biology of steroid sulfatase has been reviewed [55] as has the role of steroid sulfatase in oestrogen metabolism [56].
The gene has been cloned and sequenced, the protein expressed and purified [57,58] and shown to consist of 583 residues (Fig. 7). The structure of the protein has been solved (PDB code 1P49: [59]) and shown to have a globular head (approximately 65Å × 50Å × 45Å) and a hydrophobic tail about 40Å long (Fig. 8). This tail is embedded in the lumenal membrane of the endoplasmic reticulum. The crystal structure shows that several residues undergo post-translational glycosylation. Another posttranslational modification is that of C75 to formylglycine (FG) ( Fig. 9) [60]. This is further modified, by hydration, to form the gemdiol hydroxyformylglycine (HFG), though the resting state of the enzyme appears to be hydroxyformylglycine sulfate (HFGS) with the sulfate linked to the pro-(S) hydroxyl group.
Visual inspection of the structure shows that a number of hydrogen bonds help position the HFGS and, specifically, the side chain in the binding site (Fig. 10). The HFGS backbone carbonyl may be involved in a hydrogen bond with the backbone NH of R79, and the HFGS side chain hydroxyl may form a hydrogen bond to the side chain of K134 and H136. There are four or, possibly, five interactions of the sulfate with the protein that may serve to position the sulfate during the reaction. The first of these four is an internal hydrogen bond from the HFGS backbone NH, and the other four involve the side chains of H290, K134 and, possibly, K368. A calcium ion in the binding site has an interaction with one of the HFGS sulfate oxygens. The calcium ion also has charge-charge interactions with the side chains of D35, D36, D342 and Q343. The substrate binding site is formed by residues L74, HFGS75, R98, T99, G100, V101, L103, L164, T165, L167, R168, V177, F178, T180, G181, H290, T291, H346, E349, K368, N447, T484, H485, V486, F488 and C489.
One proposed mechanism for the steroid sulfatase-catalysed reaction is shown in Fig. 11 [61]. In Step I FG75 is activated by a water molecule to form the gem-diol. In step II the substrate sulphur undergoes nucleophilic attack by the pro-(S) hydroxyl following its activation by Ca 2+ . This results in the sulfate moiety being covalently linked to FG75 and the unconjugated substrate being released with the sulfate being replaced by a hydrogen abstracted from H290. In step III the pro-(R) hydroxyl of HFGS75 is deprotonated by H136 and the ester bond is broken by attack of the activated free hydroxyl. The sulfate is released and FG75 is regenerated. Release of the sulfate may require an additional water molecule as a nucleophile. This water may accompany the incoming substrate, shielding the charged sulfate during its passage into the substrate binding site.
The reaction site lies at the base of a long narrow pocket that opens beside the hydrophobic tail. This suggests that the substrate and product may have to travel through the endoplasmic reticulum membrane in order to enter and leave the substrate binding site. The binding site itself is formed predominantly from hydrophobic residues (Fig. 12). The crystal structure has two water molecules in the substrate binding site, one of them putatively being the catalytic water, but the other being a crystallographic water interacting with Adapted from [52]. the sulfate of HFGS75. This second water has to be removed prior to performing docking experiments to determine how the substrate and product might fit into the binding site.
Inhibition of steroid sulfatase (together with the inhibition of aromatase -see below) should reduce the amount of oestrogen in the body, particularly in target tumour tissue, and be beneficial in a range of hormone-dependent diseases. The development of steroid sulfatase inhibitors has been the subject of a number of recent reviews [62][63][64][65][66].

17␤-Hydroxysteroid dehydrogenases
There are fourteen different vertebrate enzymes classified as 17␤-hydroxysteroid dehydrogenases of which twelve have been found in human tissues [67]. Despite their name the preferred substrate of some of these enzymes is other than steroids and, when the substrate is a steroid, the reaction may be either an oxidation or a reduction depending on the cofactor and cellular localisation [68]. 17␤-Hydroxysteroid dehydrogenase type 1 (EC 1.1.1.62, 17␤-HSD1) catalyses the conversion of estrone to estradiol, and of 16␣-hydroxyestrone to estriol (Fig. 1). The reverse reactions are catalysed by 17␤-HSD2 which also facilitates the conversion of testosterone to androstenedione, the reverse reaction of which is catalysed by 17␤-HSD3.
All three enzymes are members of the short-chain dehydrogenase/reductase (SDR) structural family. The aligned sequences are shown in Fig. 13. Listed in Table 2 are the twenty-two publicly available structures of 17␤-HSD1. Most of these structures have a ligand in the substrate binding site: the structures of these ligands are shown in Fig. 14. The gene can be translated into 328 amino acids but the mature protein has lost the initiating methionine and, for historical reasons, is often (but not always) numbered A1-Q327. Herein the A1-Q327 numbering is used. There are no    publicly available crystal structures of either 17␤-HSD2 or 17␤-HSD3, but the overall 3D structure of both is believed to be similar to that of 17␤-HSD1 (Fig. 15). The protein is built around a core of seven parallel ␤-strands with at least one helix between successive strands. This makes up the Rossman fold typical of nucleotide binding proteins. The loop between sheet 1 and helix 1 has the GxxxGxG motif (residues G9-G15, where G is glycine and x any other residue) that is common to oxidation/reduction enzymes that bind nicotinamide cofactors. S142, at the end of sheet 5, and Y155 and K159, in helix 9, have been shown to be essential for activity [81]. Fig. 11. Proposed catalytic mechanism of steroid sulfatase. Adapted from [61]. The substrate binding site is formed by residues G94, L95, L96, S142, V143, G144, M147, L149, P150, N152, Y155, C185, G186, P187, F192, M193, V196, Y218, H221, S222, V225, F226, F259, L262, M279, E282 and V283. Of these twenty-six residues only fourteen are in secondary structural elements, i.e., helices or sheets, with the remainder being in loops between the structural elements. One of these loops, the substrate binding loop comprising residues H189-V196, is flexible enough that it is observed in two conformations in the 1FDT structure. This loop probably has to move to allow the substrate into the binding site. The following residues form the NADPH binding site: T8, G9, C10, S11, S12, G13, I14, G15, R37, L64, D65, V66, R67, C89, N90, A91, G92, L93, V113, T140, G141, S142, Y155, K159, C185, V188, H189, T190, A191, F192, M193 and K195.
The hydride transfer reactions catalysed by these enzymes are intrinsically reversible but, in vivo, are effectively unidirectional due to the relative concentrations of NADPH and NADP + [82]. Based on the measurement of isotope exchange rates between substrate-product pairs the reaction catalysed by 17␤-HSD1 has been reported to be a random order bi-bi mechanism [83] (Fig. 16). However, a dynamics study based on crystal structures of the apo enzyme and binary and ternary complexes has suggested that there is a preferred order with NADPH binding before the substrate, and the NADP + being released before the product [84]. In the reaction the pro-(S) hydride is transferred from NADPH to the alpha face of the estrone C17. A hydride is then transferred to the C17 oxygen, stabilised by interaction with the hydroxyl of S142, from the hydroxyl of Y155. There is then postulated to be a hydride transfer network involving the NADPH ribose hydroxyl, the K159 amino moiety and a water molecule that results in the formation of a hydronium ion [84] (Fig. 17). Work with the 17␤-HSD from the filamentous fungus Cochliobolus lunatus has shown that cofactor dissociation is the slowest step of the reaction and that the catalytic activity might be modulated by differences in the conformation of the substrate binding loop consequent upon cofactor and substrate binding [85]. Enzymes of the SDR family are typically active as dimers or tetramers, and work with the C. lunatus enzyme suggests that dimerisation may be necessary for activity [86]. Of the estrogens estradiol is the most potent and the inhibition of estradiol production is beneficial in the treatment of oestrogen-dependent diseases. Inhibition of 17␤-HSD1 should prevent estradiol production, though the concomitant inhibition of aromatase (see below) and steroid sulfatase is probably necessary for greater reduction of estradiol levels. A variety of steroidal and non-steroidal inhibitors have been synthesised that have IC 50 values in the low nanomolar range. For recent reviews of these enzymes and their inhibitors see [87][88][89].

Aromatase (cytochrome P450 19A1, oestrogen synthase)
Aromatase (cytochrome P450 19A1, cytochrome P450AROM, oestrogen synthase) catalyzes the aromatisation of the 'A' ring of androstendione to produce estrone and the aromatisation of the 'A' ring of testosterone to produce estradiol (Fig. 1). The enzyme consists of 503 amino acids (Fig. 18) and an iron-containing haem group (protoporphyrin IX; Fig. 19). The protein is glycosylated on N12 [90]. Other post-translational modifications are performed by the tyrosine kinases c-Src [91] and PTP1B [92]: phosphorylation of Y361 by c-Src increases aromatase activity. The enzyme is found in the cytoplasm bound to the endoplasmic reticulum with the Nterminal residues probably penetrating through the membrane and the glycosylated residue in the lumen of the endoplasmic reticulum. However, it has been suggested that the N-terminal residues are required not so much for membrane association as for orientation [93]. There is evidence to suggest that aromatase forms dimers or higher oligomers [93,94]. The proposed mode of oligomerisation may be a further means of regulating enzyme activity as it would prevent the phosphorylation of Y361 [94].
There is a hydrogen bond between the backbone NH of M374 and the ligand carbonyl at the 17-position and another between the side chain of D309 and the ligand carbonyl at the 3-position. In the 3EQM and 3S79 structures the androstenedione has an identical pose, as do the haems and the residues forming the binding site. The B, C and D rings of exemestane overlay the equivalent rings in androstenedione but, because of the additional double bond in the A ring of exemestane, the A ring is slightly distorted relative to the androstenedione A ring. The C6-methylidene carbon is accommodated by slight shifts (less than 1Å) in the position of T310 and, to a lesser extent, S478. Despite the greater size of the 6-substituent in the 4GL5 and 4GL7 structures there is no further significant movement of any of the amino acid residues in the substrate binding site.
The aromatisation occurs in a three-step mechanism [98,99] (Fig. 21). Two consecutive hydroxylations generate, first, a hydroxyl at C19 and, second, a gem-diol intermediate that dehydrates to leave an aldehyde [100]. The ketone at the 3-position then enolises and the aldehyde is cleaved from C10 and ejected as methanoic (formic) acid [101]. This implies that the aromatase catalytic centre can catalyse two different types of reaction (Eqs. (1) and (2)). Note that the first hydroxylation of C19, whilst leaving the absolute configuration of C10 unchanged, does change the stereochemistry of C10 from R to S. (1) (2)  13-methyl-7,8,9,11,12,13,14,15,16,17-decahydro-6H-cyclopenta[a]phenanthren-16-yl)methyl)benzamide. Fig. 15. The structure of 17␤-HSD1. A cartoon of the 17␤-HSD1 structure coloured blue at the N-terminus through to red at the C-terminus. The secondary structure elements are labelled as in Fig. 13: H -helix and S -sheet. NADPH is shown with purple carbons and estradiol with pink carbons (based on the 1A27 structure [69]). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Fig. 16. The proposed random order bi-bi mechanism of 17␤-HSD1. The mechanism believed to be preferred is shown in bold [83,84]. E -enzyme, E1 -estrone, E2 -estradiol.  Residues involved in the catalytic act probably include A306, D309 and T310. The binding site opens to the ER membrane so the substrate has to traverse the membrane in order to enter the binding site. NADPH does not bind directly to aromatase. Two electrons are transferred from NADPH to the aromatase haem group via NADPH cytochrome P450 reductase, an FAD/FMN-containing protein [102,103].
Atomic force microscopy has been used to observe aromatase dimers in membranes, and possibly higher order interactions [93]. Computational modelling of aromatase in membranes suggests that a dynamic quaternary organisation together with fluctuations of the active site, the cavity enclosing the haem and the substrate access channel are necessary for activity [104,105].
The inhibition of aromatase is a suitable treatment for a number of clinical conditions that are caused or aggravated by the overproduction of oestrogen. Aromatase inhibitors such as anastrozole, letrozole and exemestane (Fig. 22) have found roles in the treatment of breast cancer [106], ovarian cancer [107] and growth maximisation in puberty in children with short stature [108], and it has been suggested that aromatase inhibitors may have a role to play in treating some cases of lung cancer [109].

Enzymes of steroid hydroxylation
Aromatase, cytochrome P450 19A1, catalyzes the hydroxylation of testosterone and androstenedione at C19 prior to aromatising the A-ring. Other cytochrome P450s catalyse hydroxylations at other sites [110], most commonly the 2-, 4-and 16-positions. It has been argued that which cytochrome is involved with which hydroxylation is dependent on both the body tissue and the substrate concentration [111]. Cytochrome P450s 1A1, 2C19, 3A4 and 3A5 are amongst those reported to catalyse hydroxylation at the 16␣-position [111,112]. This generates 16␣-hydroxyestrone which can then be converted to estriol by 17␤-HSD1. Reported to be involved in both 2-and 4hydroxylations are cytochrome P450s 1A1, 1A2 and 1B1, though the relative importance of these might depend upon the tissue, and others are probably involved [111]. In the PDB are the structures of more than 500 cytochrome P450s, over 100 of which are of human proteins. The structures of CYPs 1A2, 1B1 and 3A4 are briefly discussed herein.
There is one crystal structure of human CYP 1A2, PDB code 2HI4 [113], and one of CYP 1B1, PDB code 3PM0 [114]. The structure of both is based on that of the canonical P450 fold -twelve ␣helices and four ␤-sheets, designated A-L and 1-4, respectivelywith a few additional helices. Both have the same haem group, protoporphyrin IX, as that in aromatase (Fig. 19). In both enzymes the substrate binding site is a narrow pocket that tightly encloses a molecule of ␣-naphthoflavone, with many of the residues forming the binding sites being the same in the two proteins. In the following listing of these residues the numbers for CYP 1A2 are given first, with the numbers for the equivalent residues in CYP 1B1 in brackets. One face of the ligand is bordered by I117 (V126), T118 (S127), S122 (S131), T124 (A133), F125 (F134), T223 (N228), F226 (F231), V227 (G232), F256 (L264), N257 (N265), L382 (V395), T385 (T398), I386 (I399), L497 (L509) and T498 (T510). F226 (F231) forms a Since both enzymes can hydroxylate estrone at either the 2or 4-position there must be sufficient flexibility in the substrate binding site, despite the restrictiveness observed in the crystal structures, to allow the substrate to bind in different poses to present either the 2-or 4-positions to the catalytic centre. That such flexibility of substrate binding exists is demonstrated by the fact that the pose of the ␣-naphthoflavone in one structure is the opposite way round to that in the other structure (Fig. 23).
There are nine publicly available structures of CYP 3A4 (Table 3). This enzyme catalyzes hydroxylation at the 16␣-position. In the various structures there are seven different structurally diverse Table 3 The crystal structures of cytochrome P450 3A4. The structures of the ligands are shown in Fig. 22 Bromoergocryptine [120] a The progesterone is not in the substrate binding site, but in a groove on the surface of the protein.
b There are two molecules of ketoconazole in the substrate binding site.   ligands (Fig. 24), though one of these, progesterone, is not in the substrate binding site but in a groove on the surface. To accommodate these ligands there is considerable variability in the shape of the substrate binding site (Fig. 25). This variability shows that there is plenty of space to accommodate estrone such that it presents the 16-position for hydroxylation.
COMT exists in two forms, soluble (s-COMT, 221 residues) and membrane bound (mb-COMT, 271 residues) that differ only in the presence of an additional fifty residues at the N-terminal end of the membrane-bound form (Fig. 26). The residue numbering for s-COMT, 1-221, will be used herein. Residue 108 is a site of a common polymorphism: it can be either valine or methionine. The rarer methionine polymorphism has been associated with a variety of neuropsychiatric disorders [125], increased risk of breast cancer [126,127], lower pain threshold [128] and improved cognitive function involving working memory [129]. These different effects of the polymorphs may be due to the slightly greater enzyme activity of the valine variant [130][131][132] and/or the greater range of conformational states of the methionine variant observed in molecular dynamics simulations [133] that were suggested by the greater instability (with respect to temperature, autoxidation and denaturants) of the methionine variant (see, e.g., [134] and references therein).
There are three crystal structures of human s-COMT, one of which is of the methionine variant (3BWY; [135]) and the others of the valine variant (3BWM; [136]: 3A7E; unpublished). All three have the methyl group donor, S-adenosyl methionine (SAM), and an inhibitor, 3,5-dinitrocatechol (DNC), bound (Fig. 27). The structure of s-COMT is one of two sets of ␣-helices sandwiching a seven-stranded ␤-sheet core (Fig. 28). A shallow cleft on the protein surface is occupied by SAM, DNC and a magnesium ion. The SAM binding pocket has a base formed by the C-terminal ends of ␤-sheets a-d with the walls of the pocket formed by helix 6 and the loops between helix 2 and helix 3, ␤-sheet a and helix 4, and ␤sheet d and helix 7. The SAM methyl group is directed towards the substrate binding site and the catechol oxygen to be methylated.
The methylation reaction is dependent on an initial strict sequential order of reactant binding. SAM binding is followed by the binding of a magnesium ion which is followed by the catechol entering the binding site [130,136]. This order is necessary because the binding of one changes the shape of the binding site so that the next can bind. If reactant binding does not occur in this order a dead-end complex is formed. The magnesium ion is held in place by coordination with three acidic residues, D141, D169 and E199 the latter two of which are oriented by interaction with K46 (Fig. 29). The oxygen acting as the methyl recipient is activated by K144 which acts as a general base in the methylation. In a direct bimolecular transfer the SAM methyl group is transferred from the sulphur to the catechol hydroxyl oxygen through an S N 2-like transition state [137].
COMT inhibitors have been developed primarily for the treatment of Parkinson's disease [138,139].

Discussion
Sulfotransferase, sulfatase, 17␤-HSD1 and aromatase all have to recognise estrone (as either a substrate or a product of the  catalysed reaction) and at least one other steroid. Despite this common recognition requirement there is no structural similarity between the four proteins in terms of the secondary and tertiary structure and little similarity in the structure of the oestrogen Fig. 28. The structure of s-COMT. A cartoon of the s-COMT structure coloured blue at the N-terminus through to red at the C-terminus. The secondary structure elements are labelled as in Fig. 27: H -helix and S -sheet. SAM and DNC are shown with purple carbons. The yellow sphere is a magnesium ion. Based on the 3BWM structure [136].
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) binding sites (Fig. 30). The possible exception to this assertion is, perhaps unsurprisingly given that they catalyse opposite reactions, sulfatase and sulfotransferase. These two enzymes both have a flat surface against which the substrate A and B rings can nestle. This flat surface is formed by L103 in sulfatase and V145 in sulfotransferase (Fig. 31). Both enzymes also have a histidine and two lysines adjacent to the cofactor sulfate group -H136, K134 and K368 in sulfatase, and H107, K47 and K105 in sulfotransferase -possibly suggesting some degree of similarity in the mechanism of the sulfate transfer reactions. Since estrogens have an important role to play in hormonedependent cancers the inhibition of the enzymes responsible for producing estrogens is equally important. Historically, most effort has been focused on the development of aromatase and sulfatase inhibitors [140], but more recently considerable effort has been put into the development of 17␤-HSD1 inhibitors [87][88][89]. Aromatase is the only enzyme capable of aromatising the 'A' ring thus converting androgens to estrogens. The inhibition of aromatase will, therefore, prevent oestrogen production. However, stores of estrone sulfate (accumulated from the circulation throughout the body at higher levels in breast cancer tissue than normal tissue [141,142]) can be converted to estrone by sulfatase and then to estradiol by 17␤-HSD1. With levels of sulfatase and 17␤-HSD1 being higher in cancerous tissue than non-cancerous tissue this has been reported to be the main route of oestrogen production in cancerous breast tissue [143-145; see, also, 146] which might suggest that the inhibition of aromatase alone would be unlikely to be a successful treatment in more than a minority of cases. However, aromatase inhibitors such as letrozole, anastrozole and exemestane have proven to be highly effective treatments for many breast cancers [106].  The different binding site structures of the various oestrogen metabolising enzymes mean that specific inhibitors can be designed. A more recent approach, however, is to design molecules that inhibit multiple drug targets, and the first reports of single molecule dual aromatase-sulfatase inhibitors have been published [147][148][149]. Inhibitors of other combinations of enzymes are feasible and the development of these will be facilitated by structure-based design. Inhibition of aromatase already reduces circulating levels of estradiol to below detectable levels. The concurrent inhibition of sulfatase and also, perhaps, 17␤-HSD1 should cut the localised in situ production of estradiol in tumour tissue. By inhibiting aromatase, sulfatase and 17␤-HSD1 all oestrogen production should be stopped. This may, however, have adverse consequences as estrogens are known to have protective effects on the heart [150]: the incidence of cardiovascular disease in postmenopausal women is three times that of premenopausal women of the same age [151]. This is due, at least in part, to the increased level of aldosterone [152] which is synthesised by CYP11B2 (aldosterone synthase). The simultaneous inhibition of CYP19 (aromatase) and CYP11B2 should reduce the levels of both oestrogen and aldosterone thus having the desired anti-cancer and anti-cardiovascular disease effects. The first dual inhibitors of CYP19 and CYP11B2 have recently been reported [8].