Preparation of 16β-Estradiol Derivative Libraries as Bisubstrate Inhibitors of 17β-Hydroxysteroid Dehydrogenase Type 1 Using the Multidetachable Sulfamate Linker

Combinatorial chemistry is a powerful tool used to rapidly generate a large number of potentially biologically active compounds. In our goal to develop bisubstrate inhibitors of 17β-hydroxysteroid dehydrogenase type 1 (17β-HSD1) that interact with both the substrate (estrone or estradiol) and the cofactor (NAD(P)H) binding sites, we used parallel solid-phase synthesis to prepare three libraries of 16β-estradiol derivatives with two or three levels of molecular diversity. From estrone, we first synthesized a sulfamate precursor that we loaded on trityl chloride resin using the efficient multidetachable sulfamate linker strategy recently developed in our laboratory. We then introduced molecular diversity [one or two amino acid(s) followed by a carboxylic acid] on steroid nucleus by Fmoc peptide chemistry. Finally, after a nucleophilic cleavage, libraries of 30, 63 and 25 estradiol derivatives were provided. A library of 30 sulfamoylated estradiol derivatives was also generated by acidic cleavage and its members were screened for inhibition of steroid sulfatase. Biological evaluation on homogenated HEK-293 cells overexpressing 17β-HSD1 of the estradiol derivatives carrying different oligoamide-type chains at C-16 first revealed that three levels of molecular diversity (a spacer of two amino acids) were necessary to interact with the adenosine part of the cofactor binding site. Second, the best inhibition was obtained when hydrophobic residues (phenylalanine) were used as building blocks.


Introduction
17β-Hydroxysteroid dehydrogenase type 1 (17β-HSD1), also called human estradiol dehydrogenase [E.C.1.1.1.62], is a protein comprised of 327 amino acids that exists as a homodimer of 35 kDa [1]. This very important steroidogenic enzyme catalyzes the last step in the biosynthesis of estradiol (E 2 ), the most potent estrogen [2][3][4]. For this purpose, this enzyme uses the cofactor NAD(P)H to stereoselectively reduce the C17 ketone of estrone (E 1 ) into a 17β-alcohol, as illustrated in Figure 1. Because E 2 concentration has been found to be significantly higher in breast tumours than in normal breast tissue [5,6] 17β-HSD1 is an interesting therapeutic target for estrogen-sensitive diseases such as breast cancer.  Although 17β-HSD1 activity was reported 50 years ago [7,8], no significant progress was made to obtain active inhibitors as drugs [9][10][11][12][13][14]. Among the known inhibitors [9] the bisubstrate category is promising because these inhibitors interact with both the substrate (E 1 ) and the cofactor (NAD(P)H) binding sites. In collaboration with Dr. S.X. Lin, our research group developed the first bisubstrate inhibitors of 17β-HSD1 by linking E 2 and adenosine with an alkyl side-chain spacer. With a spacer of eight methylenes, EM-1745 (1a, Figure 2) was found to be the best bisubstrate inhibitor of that series [15,16]. EM-1745 (1a) was also co-crystallized with the enzyme and X-ray analysis revealed that 1a interacts with both the substrate (E 2 ) and the cofactor (adenosine part of NADH) enzyme binding sites [17]. The amine and the alcohol groups on the adenosine moiety of 1a form hydrogen bonds with Asp65 and Ser11, respectively while the steroid nucleus forms hydrogen bonds with His221, Ser142 and Tyr155. Simplified bisubstrate inhibitors were also developed by our research group in order to improve the stability and the bioavailability of 1a [18,19]. The carboxylic acid derivatives 1b were found to be the best inhibitors of that series. In order to explain the high potential of compound 1b as inhibitors of 17β-HSD1, we hypothesized that, as a bioisostere of a phosphate, the carboxylic acid functional group should form hydrogen bonds with one or more of the same amino acids interacting with the phosphate of NADPH (Ser11, Arg37 and Lys195) [19][20][21]. More recently, a new generation of hybrid inhibitors of 17β-HSD1 was reported by Sterix [22][23][24] and by us [25]. Lawrence et al. [22][23][24] have reported potent inhibitors of 17β-HSD1 structurally composed of a pyridyl moiety linked by an amide at C16 position of E 1 (compounds 2a). Using the GOLD docking program, they determine that the steroid moiety interacts with the substrate-binding site while the pyridyl moiety forms interactions with the nicotinamide part of the cofactor. On the other hand, we reported a new family of 16β-E 2 derivatives (compounds 2b) [25] having a m-carbamoylbenzyl moiety that interacts with the enzyme as demonstrated by the 3D analysis of the ternary (17β-HSD1/2b/NADP) complex [26]. Taken together, the results discussed above indicate that the enzyme contains important amino acids, which could be targeted with an appropriate side-chain added at position 16β of E 2 . In order to improve the affinity of the bisubstrate inhibitors 1a and 1b for 17β-HSD1, a series of E 2 derivatives was designed. Thus, the 16β-methylene side-chain of bisubstrate inhibitor 1 was replaced by functional groups that potentially interact with the left part (nicotinamide and spacer) of the cofactor-binding site. On the other hand, an amine and/or a carboxylic acid as functional group to form interactions with the right part (adenosine) of the cofactor-binding site was added at the spacer end. In fact, the amine group on the adenosine moiety of 1a and the carboxylic acid function on 1b were found to be important for the interactions with the cofactor-binding site of this enzyme. To rapidly provide a large number of 16β-E 2 derivatives of general structure 3 for structure activity relationship (SAR) study, parallel solidphase synthesis was used. Among the resins and the linkers known for preparing libraries of E 2 derivatives by solid-phase chemistry [27] the linker sulfamate on trityl chloride resin developed by our research group was chosen [28,29]. This multidetachable type linker can generate a phenol or a sulfamate derivative depending on cleavage conditions as presented in Figure 3. It could be used for the synthesis of steroidal and non-steroidal compounds, thus adding to its potential. Furthermore, this linker has already been used to successfully prepare several libraries of E 2 derivatives, sulfamoylated or not [30][31][32]. The members of these libraries provided 17β-HSD1 inhibitors or steroid sulfatase potential inhibitors, respectively [32]. We report herein the preparation of one library of sulfamoylated E 2 derivatives (library A: 30 members) and three libraries of E 2 derivatives (libraries B, C and D: 30, 63 and 25 members, respectively). Biological evaluation of members of libraries B and C on homogenated HEK-293 cells overexpressing 17β-HSD1 will also be presented and discussed. Finally, the chemical problems met during the preparation of library D will be discussed and solutions will be proposed.

Results and Discussion
New bisubstrate inhibitors of 17β-HSD1 that should interact with the substrate binding site and the left (nicotinamide and spacer) and the right (adenosine) part of the cofactor binding site were designed and prepared by parallel solid-phase synthesis. One or two amino acid building blocks were introduced at position C16β of the E 2 scaffold linked on solid support by the sulfamate linker, followed by the addition of a carboxylic acid building block. After removal of the protective group and nucleophilic cleavage, peptidosteroids with the general structure 3 were generated. The preparation of peptidosteroids was chosen since side chain of natural and non-natural amino acids provides an interesting pool of functionalities. Furthermore, amino acid coupling procedures are well known for solid-phase synthesis and the Fmoc coupling strategy could be easily used with the sulfamate linker.

Synthesis of steroidal solid-phase precursors 6a and 6b in solution (Scheme 1)
Before the preparation of libraries A-D, compounds 6a and 6b were synthesized in solution as presented in Scheme 1. Azide intermediates 4a and 4b were prepared as reported previously in 8 and 9 steps, respectively, from commercially available E 1 [33]. Since amino acids will be introduced using the Fmoc-coupling procedure, a Fmoc on the aminopropyl side chain at position C16β was selected as an appropriate protective group. For this purpose, azides 4a and 4b were first reduced using hydrogen catalyzed by Pd/C. The C3 silylated protective group was next removed using TBAF in THF. Compounds 5a and 5b were subsequently obtained by selective protection of the amine with a Fmoc group. Treatment of 5a and 5b in DCM with sulfamoyl chloride [34] in the presence of a non-nucleophilic base, 2,6-di-tbutyl-4-methylpyridine, provided solid-phase precursors 6a and 6b in 4 and 3 steps (82% and 52% overall yield, respectively) from 4. Scheme 1. Synthesis of solid-phase precursors 6a and 6b.  and 5b); (e) NH 2 SO 2 Cl, 2,6-di-t-butyl-4-methylpyridine, DCM, rt, 16-24 h (6a: 82% for 4 steps; 6b: 52% for 3 steps).

Synthesis of carboxylic acid building blocks 14-18 (Scheme 2)
For library C members, capping with an amine functional group was chosen to interact with the cofactor (adenosine) binding site. For this purpose, aniline derivatives were chosen. In order to obtain optimal interactions with the cofactor-binding site of the enzyme, carboxylic acid with several alkyl spacer lengths (n = 0 to 3 methylenes) was chosen. Aniline derivatives 9-10 with a spacer of two methylenes were not commercially available, but were prepared easily in one step from 4-aminocinnamic acid or 3-nitrocinnamic acid, respectively, as previously reported [35]. In order to avoid polymerisation during the capping coupling step on solid-phase organic synthesis, free anilines 7-11 were protected as Fmoc using FmocOSu and NaHCO 3 in a mixture THF/H 2 O (5:1) to provide 14-18 in good yields (51-93%). It is noteworthy to mention that Fmoc-aniline derivatives 12-13 were commercially available.

Solid-phase synthesis of libraries A, B and C (Scheme 3)
A library of 30 sulfamoylated E 2 derivatives (A), a library of 30 E 2 derivatives (B), and a library of 63 E 2 derivatives (C) were prepared by parallel solid-phase synthesis using the multidetachable linker sulfamate. Precursor 6a was first loaded on trityl chloride resin. For this reaction, trityl chloride resin was swelled in dry DCM and treated with 6a and diisopropylethylamine (DIPEA) in a peptide flask. After 16 h of shaking, the reaction mixture was filtered and washed with DCM and MeOH to obtain resin 19. The loading yield of 19 was calculated by the increase of the resin weight. This yield was 70% for libraries A and B and 42% for library C. A lower loading yield was obtained for library C because 1 equivalent of 6a was used for 2 equivalents of trityl chloride resin instead of 1 equivalent of resin used in the preparation of libraries A and B. On a model library with a loading of 75%, completion of the coupling reaction was very difficult for the introduction of the third level of molecular diversity. It was hypothesized that steric hindrance could be responsible for the lower reactivity of the amine on the steroid. Therefore, less precursor 6a was loaded on resin when more than two levels of molecular diversity needed to be introduced on the steroid. In the next step, resin 19 was treated for 1 h with a freshly prepared solution of 20% piperidine in DCM to remove the Fmoc protective group and to free the amine for the next step. It is noteworthy to mention that after each solid-phase organic step, the resin was washed with the appropriate solvent and dried under a vacuum. Furthermore, the solid-phase reactions were monitored by a minicleavage test of a random sampling of resin with 5% TFA in DCM.
The resin 20 was next split into 30 equal portions for libraries A and B and 63 equal portions for library C. The resins were then placed in bottom fritted reaction vessels of a 96 solid-phase reaction block of an ACT-Labtech semi-automated synthesizer. The first level of molecular diversity (R i ) was introduced on each resin 20 with one of a selection of Fmoc-protected amino acids from L series. Fmoc-L-Phe-OH, Fmoc-L-Ile-OH, Fmoc-L-Pro-OH, Fmoc-L-Ala-OH and Fmoc-Gly-OH were used as building blocks for libraries A and B, and Fmoc-L-Phe-OH, Fmoc-L-4-NO 2 -Phe-OH and Fmoc-Gly-OH were chosen as building blocks for library C. The coupling reaction was performed for 2 to 3 h with DIPEA in DMF using PyBOP and HOBt as coupling reagents. The Fmoc protective group was then removed as described above to afford resins 22 from 21. Introduction of a second level of molecular diversity (n = 2) on resins 22 using the same procedure and Fmoc-L-amino acids (Phe, 4-NO 2 -Phe and Gly) was carried on for library C. Removal of Fmoc group then provided resins 24 from 23. For members of both libraries, a second or a third level of molecular diversity (R j ) was introduced using a selection of carboxylic acids as building blocks (4-pyridylacetic acid, benzoic acid, acid 14, 2-pyrazinecarboxylic acid, 3-aminopyrazine-2carboxylic acid, and indole-2-carboxylic acid for libraries A and B, and carboxylic acids 12-18 for library C). These carboxylic acids were preactivated with PyBrOP and HOBt. The reaction was carried on for 2 to 3 h with DIPEA in DMF. Then, resins 25 (n = 1 or 2) that were prepared with carboxylic acids 12-18 as building blocks were treated with a solution of 20% piperidine in DCM in order to remove the Fmoc protective group. Finally, the C17-acetate protective group of resins 26 and 27 was removed by treatment with MeONa in MeOH and THF to afford resins 28.
The last step was to release the phenolic and the sulfamate derivatives from the solid support by either a nucleophilic or an acid treatment, respectively. The nucleophilic cleavage was done first because the phenolic derivatives are not totally released from the solid support under these conditions. Furthermore, E 2 derivatives are preferred since building blocks were chosen to target 17β-HSD1. Thus, resins 28 were treated with a solution of 30% diethylamine (DEA) in THF at room temperature for 24 to 66 h to release E 2 derivatives 59-88 (library B) and 89-151 (library C). After the nucleophilic treatment, the resin was washed with acetone or THF and MeOH for libraries B and C, respectively. The filtrate was evaporated. The crude product was dissolved in EtOAc or THF and evaporated again. This procedure was done twice in order to provide DEA-free E 2 derivatives. The acid cleavage was performed on resins 28 that were previously treated by a nucleophilic cleavage for 24 h to generate library B. In fact, there were still steroids loaded on resins 28 after the nucleophilic cleavage. Thus, these resins 28 were treated with a solution of 30% hexafluoroisopropanol (HFIP) in DCM for 6 h at room temperature. Evaporation of the solvent afforded sulfamoylated E 2 derivatives 29-58 (library A). The acid cleavage was not carried on with 5% TFA in DCM to avoid the formation of C17 trifluoroacetate derivative as previously reported in the preparation of a 17α-substituted E 2 sulfamate library [31].

Characterization of libraries A, B and C (Tables 1-3)
Sulfamoylated E 2 derivatives of library A and E 2 derivatives of libraries B and C were submitted to a random sampling after we confirmed the presence of a major compound by TLC analysis. Six members of library A, six members of library B and 14 members of library C were then characterized by 1 H-NMR and LRMS. In addition, the purity of the sampling of libraries A and B was determined by HPLC analysis. Expected masses matched for all these compounds. Furthermore, 1 H-NMR analysis confirmed the structure of the library members. The HPLC-purity of final compounds ranged from 89 to 91% and 85 to 95% for libraries A and B, respectively. The average HPLC-purity was 90% for both libraries. Results were very satisfactory because no purification step was performed after the final cleavage. The excess of reagent (HFIP and DEA) and solvent were simply removed by evaporation. The average overall yields for the solid-phase sequence of reactions (6 or 7 steps for libraries A and B, and 9 steps for library C) were 46%, 27% and 31%, respectively. The average overall yield was calculated after the loading step of precursor 6a or 6b on trityl chloride resin. The mass of phenols released from solid-support by nucleophilic cleavage to provide library B was not removed in the calculation of the overall yield of library A. The purity of library C members was estimated to be around 70-75% by TLC and 1 H-NMR analysis. The lower purity of library C members could be explained by the final deprotection step on solid-phase. In fact, resins 27 were treated with a solution of MeONa in MeOH and THF for approximately 100 h instead of 48 h in order to complete the reaction. However, this long treatment resulted in degradation of part of the product. Nevertheless, the purity of library C members was found sufficiently satisfactory for screening assay on 17β-HSD1.

Chemical synthesis of compound 156 used in the preparation of library D (Scheme 4)
The capping groups of members from library D have a carboxylic acid as functional group to interact with the cofactor (adenosine) binding site of 17β-HSD1. In fact, among the simplified bisubstrate compounds prepared to inhibit this enzyme, carboxylic acid derivatives were found to be by far the best inhibitors of that series [36]. The introduction of the capping building block on solid-phase was done by a reaction between an amine and a carboxylic acid. Indeed, the capping synthon 156 is a dicarboxylic acid and one of the COOH group has to be appropriately protected. The tert-butyl ester was chosen as protective group because it could be easily removed in acid conditions after the nucleophilic cleavage. Thus, starting from commercially available 2-(3-bromophenyl)-acetic acid (152), the carboxylic acid was protected with di-tert-butyl dicarbonate and dimethylaminopyridine (DMAP) in tert-butanol to afford tertbutyl ester 153 in an excellent 94% yield [37]. A vinyl was next added on aryl bromide 153 by a Stille reaction using mild conditions (tributylvinyl tin, Pd 2 (dba) 3 , P(t-Bu) 3 in toluene) to afford 154 in 88% yield [38]. Hydroboration of vinyl 154 using standard condition provided alcohol 155. This alcohol was oxidized under Jones' conditions in order to give carboxylic acid 156. However, the reaction conditions were too drastic and afforded the dicarboxylic acid product instead of 156. A 2-step procedure was then necessary to avoid hydrolysis of the tert-butyl ester. Thus, alcohol 155 was first oxidized with Dess-Martin periodinane to provide the aldehyde, which was subsequently oxidized in carboxylic acid using mild conditions (NaClO 2 , NaH 2 PO 4 and 2-methyl-2-butene in t-BuOH). Thus, carboxylic acid 156 was synthesized in five steps and 26% overall yield.

Solid-phase synthesis of library D (Scheme 5)
The preparation of library D required a series of Fmoc-protected amino acids with a functionalised side chain as building blocks instead of a hydrophobic one. Therefore, these amino acids needed an appropriate protective group. A tert-butyl ester and a trityl group were chosen as protective groups of the amino acid side chain since they could be easily removed in acid conditions as for the tert-butyl ester found on the capping building block. Furthermore, the molecular diversity was limited to the introduction of two amino acids. The same carboxylic acid 156 was introduced on each member of library D. Thus, focus was put on the interactions with the left part (nicotinamide and the spacer) of the cofactor-binding site. Library D was prepared as previously described for libraries B and C. Briefly, sulfamate precursor 6b was loaded on trityl chloride resin. The loading yield of 6b on the resin was 40%. Since three levels of diversity were introduced on the steroid scaffold, 1 equivalent of precursor 6b was used for 2 equivalents of trityl chloride resin as for library C. The two levels of molecular diversity were then introduced by coupling Fmoc-protected amino acids on resin 158. Fmoc-L-Phe-OH, Fmoc-L-4-Pal-OH or Fmoc-L-Ala-OH, Fmoc-L-Asp(OtBu)-OH, Fmoc-L-Ser(Trt)-OH and Fmoc-L-Tyr(2-Cl-Trt)-OH were selected as building blocks for both levels. The carboxylic acid 156 was introduced as a capping group on resins 162 in DMF and DIPEA, using PyBrOP and HOBt as coupling reagent. Removal of the C17-THP and the trityl protective group in one step on resin 163 was tried but without success. Thus, the trityl and the 2-chlorotrityl were removed by two 10-minute treatments with a solution of TFA/TIS(triisopropylsilane)/ DCM (2:1:97). The C17-THP protective group was next taken off with a solution of p-TSA in DCM and t-BuOH. Phenols were then released from the solid support under nucleophilic conditions (30% DEA/THF) for 65 h. Next, tert-butyl ester of compounds 165 was hydrolysed under acid conditions. Reaction of esters 165 with a solution of HCl (4M) in dioxane for 3 h was found to give the best conditions of hydrolysis. Evaporation of the reaction mixture provided carboxylic acids 166-190. However, these carboxylic acids were not pure enough for screening assay on 17β-HSD1. For this purpose, crude carboxylic acids (dried for at least 16 h under a vacuum) were dissolved in MeOH and preadsorbed on C-18 silica gel. They were subsequently purified by flash chromatography on C18-silica gel (reverse phase) with a mixture of H 2 O and MeOH as eluent.

Characterization of library D (Table 4) and discussion
All products were characterized by LRMS and 1 H-NMR and the results are presented in Table 4. The molecular ion peak was found in LRMS for all expected compounds (except for compounds 174 and 181) and 1 H-NMR analysis reveals the presence of the steroid nucleus and the molecular diversity. However, only six compounds were obtained as carboxylic acid derivatives. In fact, the presence of numerous peaks in the region of 3.4 to 3.7 ppm in the 1 H-NMR revealed that methyl ester contaminated the carboxylic acid products in different proportions for several compounds. Furthermore, a peak corresponding to [M+H+14] + or [M-H+14]was found in almost all LRMS spectra. We found that the methyl ester was produced during the preabsorption process on silica gel using MeOH as solvant. Even if the carboxylic acids were dried under a vacuum for 16 h after the hydrolysis step, traces of HCl trapped in the product seem to be enough to promote the formation of the methyl ester. Unfortunately, the carboxylic acid function was very important for the interactions with the cofactor (adenosine) -binding site. For this reason, carboxylic acids 166-190 were not tested as inhibitors of 17β-HSD1.
The average overall yield for the preparation of library D (9 solid-phase steps and 1 solution-phase step) was 16% after purification. The lower overall yield was generally obtained with serine derivatives, which seem to be less tolerant under acid conditions during the final deprotection step. In fact, an important proportion of a less polar compound was observed but could not be identified. The main problem during the preparation of library D occurred with the hydrolysis of the tert-butyl ester. After the nucleophilic cleavage step, the products were obtained with a good purity estimated to be higher than 80% by TLC analysis of all members and 1 H-NMR analysis of two compounds (165: R 1 = a and R 2 = e; 165: R 1 = c and R 2 = d). The choice of protective groups for this library was based on the fact that trityl and tert-butyl ester could be easily removed with TFA/H 2 O/TIS (95:2.5:2.5) and the product obtained should precipitate in diethyl ether as described in the note from the NovaBiochem catalogue. Even if this procedure works well for peptide synthesis, the acid conditions were too drastic for our steroid derivatives, resulting in decomposition of the product. Knowing this fact, a new chemical strategy could be used for the preparation of library D. The main difference would be the structure of the building block for the capping. Instead of using a tert-butyl ester to protect the carboxylic acid, a methyl ester could be used. It is obvious that aspartic acid used as a building block should also be protected as a methyl ester. Thus, the solid-phase synthesis would be very similar to the preparation of library D and the methyl ester would be hydrolysed at the last solid-phase step, using a solution of NaOH in THF. This procedure was previously described for the hydrolysis of an amino triflate loaded on solid support by the sulfamate linker. 31 Using this method, or alternatively a benzyl ester as protective group, library D members should be obtained in good purity without any purification.

Inhibition of steroid sulfatase by sulfamoylated E 2 derivatives from library A
The inhibitory potential of the members of library A on the steroid sulfatase was tested and the results were presented in our early report [30]. A sulfamate on position C3 of the steroid is an essential functional group to obtain a potent steroid sulfatase inhibition [39]. Among the 30 sulfamoylated E 2 derivatives of this library, compound 34 gave a better inhibition than estrone sulfamate (EMATE), a well known and potent inhibitor of steroid sulfatase [40]. The polar building blocks were however chosen to target 17β-HSD1 instead of steroid sulfatase, the inhibitory result could thus be better when using hydrophobic groups as elements of molecular diversity. In fact, potent inhibitors of steroid sulfatase were obtained when hydrophobic groups were added on the C16-or C17-position of E 2 sulfamate [30-32].

Inhibition of 17β-HSD1 by E 2 derivatives from libraries B and C
Inhibition of the transformation of [ 14 C]-E 1 into [ 14 C]-E 2 by members from libraries B and C was tested on homogenated human embryonic kidney (HEK)-293 cells transfected with a vector encoding for 17β-HSD1 according to a previously reported procedure [15,41]. Briefly, the enzymatic assay was conducted for 2 h at 37 °C in a sodium phosphate buffer (pH = 7.4) with NADH as cofactor.
The screening assay on 17β-HSD1 for members of library B (Table 2) is presented in Figure 4. There are two levels of molecular diversity in this library. Hydrophobic amino acids were mainly chosen for the first level to avoid the protective group on amino acid side chains. Pyridine, phenyl, aniline, pyrazine, 3aminopyrazine and indole nucleus were selected for the second level of molecular diversity. These capping building blocks were chosen to mimic the adenosine moiety found in EM-1745 (1), one of the most potent inhibitors of 17β-HSD1 [20]. For library B, the best inhibitions were obtained when phenylalanine was used as a building block for the first level with 16 to 57% at 1 µM for compounds 59-64. In general, lower inhibitions were obtained with isoleucine and proline (compounds 65-76) and almost no inhibition was observed when alanine and glycine were used as building blocks for the first level (compounds 77-88). For the second level of molecular diversity, better inhibitions were usually obtained with pyrazine, 2-aminopyrazine and indole moieties. However, these compounds weren't more potent inhibitors than unlabeled-E 1 (59% at 1 µM) and EM-251 (52% at 1 µM), a known inhibitor of 17β-HSD1 used as reference compound [42]. They were also no more potent than EM-1745 (1, 97% at 1 µM). Therefore, in order to improve the inhibitory potency of library B members, another library was prepared. The members of library C have longer spaces between the C16β E 2 and the polar capping building block. This way, the capping was able to interact with the cofactor (adenosine) binding site. Fmoc-L-Phe-OH and Fmoc-L-4-NO 2 -Phe-OH were chosen as building blocks for the first and second level of molecular diversity because members of library B with a phenylalanine as building block gave the best inhibitors. Glycine was also chosen but should give less potent inhibitors. Aniline derivatives (compounds 12-18) were selected as building blocks for the third level of molecular diversity. In fact, the aniline should form interactions with Asp65 such as the amine of the adenosine moiety of EM-1745 does in the cofactor-binding site [21]. Members of library C were tested to determine their potential to inhibit the enzyme activity at 1 µM and the results are presented in Figure 5. Three points can be highlighted from these results. First, better inhibitions were generally obtained with three levels of molecular diversity (library C) than with two levels (library B). In fact, the average percentage of inhibition for library B is 26% while for library C it is 56%. Second, the best inhibitors were obtained when phenylalanine was used as a building block for the first level of molecular diversity (R i ' = m, compounds 110-130) with more than half of the compounds giving better inhibition than EM-251 (59% at 1 µM) and eight compounds giving a better inhibition than unlabeled-E 1 (73% at 1 µM). Good inhibitions were also obtained with 4-nitrophenylalanine used as building block for the first or the second level of molecular diversity. However, as expected, lower inhibition was obtained with glycine. Third, similar inhibitions were found regardless of which aniline building blocks were selected for the third level. Globally, 26 of the 63 members of library C gave better inhibition than EM-251, and eight compounds (117, 119, 120, 121, 124, 126, 128 and 147) gave a better inhibition than E 1 .

Conclusions
We have successfully used the sulfamate linker to prepare one library of sulfamoylated E 2 derivatives and two libraries of E 2 derivatives by parallel solid-phase synthesis. Precursor 6a was loaded on trityl chloride resin via the sulfamate linker and two or three levels of molecular diversity were introduced using Fmoc-protected amino acids or carboxylic acids as building blocks. Acid (HFIP/THF) or nucleophilic (DEA/THF) cleavage respectively released the sulfamoylated E 2 derivatives (library A) or the E 2 derivatives (library B) from the resin in excellent HPLC purity (90% for libraries A and B). However, E 2 derivatives from library C were obtained with a lower purity (around 70-75%) estimated by TLC and 1 H-NMR analysis. Members of libraries B and C were also screened on homogenated HEK-293 cells overexpressing 17β-HSD1. Compounds of library C, with three levels of molecular diversity, gave better inhibition than E 2 derivatives of library B. Furthermore, 26 compounds gave a better inhibition than EM-251, a known inhibitor of 17β-HSD1 and 8 of these compounds gave a better inhibition than E 1 , used as an inhibitor. In both libraries, better inhibitors where those bearing a phenylalanine as building block. The eight top inhibitors of library C should be evaluated as pure compounds on 17β-HSD1 in order to confirm the potential of these inhibitors. However, none of the members of both libraries gave a better inhibition than EM-1745, one of the most potent inhibitors of this enzyme. In order to obtain more potent inhibitors, a third library of E 2 derivatives (library D) was also prepared as described above. However, Fmocprotected amino acids with functionalised side-chain and a protected carboxylic acid were used as building blocks. Unfortunately, the final hydrolysis of tert-butyl ester did not work as well as expected and methyl ester contaminated the final carboxylic acid products. The use of a methyl or a benzyl ester as protective group instead of a tert-butyl ester is suggested as a mean to avoid the final purification step. It is also expected that the E 2 derivatives reported in library D will more closely mimic the EM-1745/17β-HSD1 interactions than those reported in libraries A-C, thus resulting in better 17β-HSD1 inhibition.

General
Trityl chloride resin, Fmoc-OSu, coupling reagents and Fmoc amino acids were supplied by EMD Biosciences ( 1 H and 13 C respectively. Duplication of NMR signals was generally recorded for THP derivatives (presence of 2 stereoisomers). These 2 stereoisomers increased the complexity of 13 C-NMR spectra, and additional peaks are written in parentheses. Assignment of NMR signals was made easier using literature data [43]. Lowresolution mass spectra (LRMS) were recorded with an LCQ Finnigan apparatus (San Jose, CA, USA) equipped with an atmospheric pressure chemical ionisation (APCI) source on positive or negative mode. The starting material 3-(3-tert-butyldimethylsilyloxy-17β-acetoxy-estra-1,3,5(10)-trien-16β-yl)-azidopropane (4a) was synthesized as previously reported [33]. A suspension of 4a (680 mg, 1.32 mmol) and 5% Pd/C (136 mg) in EtOAc (15 mL) and MeOH (160 mL) was stirred under a hydrogen atmosphere at room temperature. After 2 h, the resulting suspension was filtered through Celite, washed with MeOH and evaporated to dryness to afford the amine (620 mg, 96%) in the form of a viscous colourless oil in good purities without purification. To a solution of the crude amine (620 mg) in anhydrous THF (40 mL) under an argon atmosphere at 0 °C was added a solution of TBAF (1M in THF, 1.53 mL, 1.53 mmol). After 30 min, the reaction was quenched by addition of a saturated aqueous solution of NaHCO 3 . The crude product was extracted with EtOAc and the organic phase was washed with brine, dried over MgSO 4 and evaporated to dryness to afford the phenol (585 mg) in the form of a yellowish solid. The crude phenol (500 mg) was dissolved in THF (120 mL) and H 2 O (40 mL). At 0 °C, NaHCO 3 (1M in H 2 O, 1.84 mL, 1.84 mmol) and Fmoc-OSu (570 mg, 1.69 mmol) were added and the reaction was stirred for 1 h at 0 °C. Then, the reaction mixture was quenched by addition of H 2 O and the crude product was extracted with EtOAc and DCM. Each organic phase was washed with brine, dried over MgSO 4 , and evaporated to dryness to provide 5a (1.05 g) in the form of a white foam. To a solution of crude phenol 5a (1.05 g) in dry DCM (120 mL) under an argon atmosphere were added 2,6-di-t-butyl-4-methylpyridine (933 mg, 4.55 mmol) and NH 2 SO 2 Cl [34] (421 mg, 3.64 mmol). The reaction was stirred for 24 h at room temperature and quenched by the addition of ice water. The crude product was extracted with DCM and the organic phase was washed with brine, dried over MgSO 4 , and evaporated to dryness. Purification by flash chromatography (hexanes/EtOAc, 6:4) gained precursor 6a (604 mg, 82% yield for 4 steps) in the form of a white foam. IR (film) 3393 (NH and NH 2 ), 1716 (C=O, ester and carbamate), 1374 and 1188 (S=O, sulfamate); 1

Synthesis of resin 20 for preparation of libraries A and B
The loading reaction was run in two 25 mL peptide flasks using the same amount of resin. In each peptide flask, a solution of precursor 6a (1.07 g, 1.62 mmol) in dry DCM (10 mL) was added to trityl chloride resin (1.09 g, 1.24 mmol/g theoretical loading) under an argon atmosphere. After 5 min, diisopropylethylamine (DIPEA, 2.16 mL) was added and the mixture was shaken for 12 h at room temperature. It is noteworthy to mention that to avoid total evaporation of DCM during the loading step, the argon inlet was removed after 30 min of reaction. The resin was filtered and washed with DCM (3×), MeOH (3×), and again with DCM, then dried overnight under a vacuum to afford 3.34 g of resin 19 globally. The loading yield, calculated by the mass increase, was 70%. The deprotection step was also run in two 25 mL peptide flasks using the same amount of resin. Resin 19 (1.67 g) was swollen in 15 mL of a solution of piperidine (20%) in DCM and shaken for 90 min at room temperature. Then, the resin was filtered and washed with DCM (3×), MeOH (3×), and again with DCM, and dried under a vacuum for 16 h. Acidic mini-cleavage (5% TFA/DCM, 10 min) of a sample of resin 20 and TLC analysis confirmed the complete deprotection of the secondary amine. The two batches of resin 20 prepared as described above were mixed before the next step. Then, 30 bottom fritted reaction vessels of the 96 solid-phase reaction block of ACT LabTech manual synthesizer were loaded with 80 mg of resin 20, which correspond to 0.052 mmol of compound 6a (without the Fmoc protective group).

Generation of E 2 derivatives (library B) by nucleophilic cleavage
To each of the 30 resins 28 was added a solution of diethylamine (DEA, 30% in THF, 1.5 mL). The mixtures were shaken for 24 h at room temperature, then filtered and washed with acetone (3×). The filtrates were collected in preweighed tubes and evaporated in a Speedvac apparatus. Each product was dissolved in EtOAc, evaporated twice, and dried under a vacuum pump in order to obtain the DEA-free product. The phenol derivatives 59-88 (2.6-18.8 mg, 10-63% overall yield from 19, Table 2

Generation of sulfamoylated E 2 derivatives (library A) by acidic cleavage
To each of the 30 resins 28 which have already been treated by nucleophilic cleavage conditions, was added 1 mL of a solution of 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) (30%) in DCM. The mixtures were shaken for 6 h at room temperature, then filtered and washed with acetone (2×). The filtrates were collected in preweighed tubes, evaporated in a Speedvac apparatus and dried under a vacuum pump. The sulfamate derivatives 29-58 (10.5-18.2 mg, 35-53% overall yield from 19, Table 1) were obtained in high average HPLC purity (90%) according to a random sampling of six library members, compounds 30, 38, 41, 49, 51 and 58. (30). Light yellow solid (15.4 mg, 45% yield); IR (film) 3307 (OH, NH and NH 2 ), 1636 (C=O, amide), 1374 and 1184 (S=O, sulfamate); 1 (38). Light yellow solid (16.2 mg, 50% yield). HPLC purity = 90%. The characterization of compound 38 was reported earlier by us [30].   -16β-(N-1H-indole-2-carbonyl-glycine-aminopropyl)-17β-hydroxy-estra-1,3,5(10 The loading procedure described above for the preparation of libraries A and B was used without any significant modifications. Briefly, trityl chloride resin (10.1 g, 1.50 mmol/g theoretical loading) and precursor 6a (5.11 g, 7.6 mmol) were put in a round 250 mL dry flask equipped with a septum and an argon inlet. DCM (100 mL) was added and the reaction mixture was stirred for 3 min. Then, DIPEA (24.3 mL) was added and the mixture was shaken at room temperature. After 30 min of reaction, the argon inlet was removed. After 16 h, the mixture was filtered, washed with DCM (3×) and MeOH (3×), and dried under a vacuum to afford 14.1 g of resin 19. The loading yield, calculated by the mass increase, was 42%. It is noteworthy to mention that the loading could not be higher than 50% because 2 equivalents of trityl chloride resin were used for 1 equivalent of precursor 6a. Then 63 fritted bottom reaction vessels of the 96 solid-phase reaction block of ACT LabTech manual synthesizer were loaded with 200 mg of resin 19. A solution of piperidine (20%) in DCM was added to resins 19 and the mixtures were shaken for 1 h at room temperature to remove the Fmoc protective group. Then, the resin was filtered and washed with DCM (2×), MeOH (3×), and dried under a vacuum for 16 h. The completion of the deprotection was proven by acidic mini-cleavage as described above. Thus, each reaction vessel contains resins loading 0.109 mmol of compound 6a (without the Fmoc protective group).

Introduction of three levels of molecular diversity (library C)
The levels of molecular diversity were introduced using a similar procedure as the one described above

Removal of protective groups (library C)
The Fmoc protective group from carboxylic acid building blocks on resins 25 (n = 2) was removed with the same procedure described above for the preparation of libraries A and B. The acetate protective group of resins 26 (n = 2) was removed using a similar procedure described above for the preparation of libraries A and B. However, the mixtures were shaken for 100 h instead of 48 h in order to obtain complete deprotection of the steroidal C17β-hydroxy by TLC analysis after acidic mini-cleavage of samples of resins 28 (n = 2).

Generation of E 2 derivatives (library C) by nucleophilic cleavage
To each of the 63 resins 28 (n = 2) was added 1.5 mL of a solution of DEA (30%) in THF. The mixtures were shaken for 50 h at room temperature. Then 1.0 mL of a solution of DEA (30%) in THF was added to each of the 63 resins and the mixtures were shaken for an additional 16 h. The mixtures were next filtered and washed with THF (1×) and MeOH (3×). The filtrates were collected in preweighed tubes and evaporated in a Speedvac apparatus. Each product was dissolved in THF, evaporated and dried under a vacuum pump in order to obtain the DEA-free product. The phenol derivatives 89-151 (5.3-33.5 mg, 6-41% overall yield from 19, Table 3) were obtained in purity around 70 to 75% according to TLC analysis of all library members and 1 H-NMR and LRMS analysis of a random sampling of 14 library members, compounds 89, 94, 99, 103, 109, 111, 114, 118, 121, 126, 133, 143, 144 and 148. 16β-(N-4-Aminobenzoyl-glycine-glycine-aminopropyl)-3,17β-dihydroxy-estra-1,3,5(10)-triene (89). Yellowish solid (16.4 mg, 27% yield); 1          To a solution of vinyl 154 (4.89 g, 22.4 mmol) dissolved in dry THF (450 mL) at 0 °C was added 1.0 M borane-THF complex (100 mL, 100 mmol) under an argon atmosphere. The mixture was stirred at 0 °C for 3 h. Then, 4 M aqueous NaOAc solutions (28.0 mL, 112 mmol) and 30% (w/v) H 2 O 2 (12.7 mL, 112 mmol) were added. After 2 h at 0 °C, the reaction was quenched by water and the extraction was performed with EtOAc. The organic phase was washed with brine, dried over MgSO 4 , and evaporated to dryness. The crude compound was purified by flash chromatography (hexanes/EtOAc, 7:3 to 6:4) to give alcohol 155 (2.36 g, 45% yield) in the form of a colourless oil. IR (film) 3435 (OH), 1732 (C=O, ester); 1  Alcohol 155 (500 mg, 2.11 mmol) was dissolved in dry DCM (42 mL) under an argon atmosphere. Dess-Martin periodinane (988 mg, 2.33 mmol) was then added and the reaction was stirred at room temperature. After 1 h, the mixture was evaporated to dryness. The crude product was filtered through a pad of silica gel using hexanes/EtOAc, 9:1 as eluent to afford the aldehyde (385 mg, 78% yield) in the form of a colourless oil. 1  . The aldehyde (372 mg, 1.59 mmol) was dissolved in t-BuOH (22 mL) and 2-methyl-2-butene (9 mL). An oxidative solution freshly prepared by dissolving NaClO 2 (1.59 g) and NaH 2 PO 4 (1.59 g) in H 2 O (15.9 mL) was added and the reaction mixture was stirred for 15 min at room temperature. The reaction was quenched by addition of water and the extraction was performed with EtOAc. The organic phase was washed with brine, dried over MgSO 4 and evaporated to dryness under reduced pressure to afford carboxylic acid 156 (350 mg, 88% yield) in the form of a colourless oil. IR (film) 3700-2300 (OH, COOH), 1731 (C=O, ester), 1713 (C=O, acide); 1  Two coupling reactions were run at the same time using a 50 mL (flask A) and a 25 mL (flask B) peptide flask. A solution of precursor 6b (2.50 g, 3.50 mmol) in dry DCM (30 mL) was prepared and 20.7 mL were added in flask A containing trityl chloride resin (3.0 g, 1.6 mmol/g theoretical loading) under an argon atmosphere. 9.3 mL of the solution of precursor 6b in DCM was added, under argon atmosphere to flask B containing trityl chloride resin (1.36 g, 1.6 mmol/g theoretical loading). After 5 min, DIPEA (7.6 mL in flask A and 3.4 mL in flask B) was added and the mixtures were shaken for 30 min at room temperature. Then, the argon inlets were removed and the mixtures were shaken for an additional 16 h. The resin was filtered and washed with DCM (3×) and MeOH (4x), and dried overnight under a vacuum to afford resin 157. The deprotection step was also run in 2 flasks, a 50 mL (flask A) and a 25 mL (flask B) peptide flask. Resin 157 was swollen in a solution of piperidine (20%) in DCM (25 mL in flask A and 10 mL in flask B) and shaken for 1 h at room temperature. The resin was then filtered and washed with DCM (3×) and MeOH (3×), and dried under a vacuum for 16 h to afford globally (flask A + flask B) 5.47 g or resin 158. Acidic mini cleavage (5% TFA/DCM, 10 min) of a sample of resin 158 and TLC analysis confirmed the complete deprotection of the amine. The loading yield, calculated by the mass increase, was 40%. It is noteworthy to mention that the loading yield could not be higher than 50% since 2 equivalents of trityl chloride resin were used for 1 equivalent of precursor 6b. Then 25 bottom fritted reaction vessels of a 40 solid-phase reaction block of ACT LabTech manual synthesizer were loaded with 200 mg of resin 158, which correspond to 0.102 mmol of compound 6b (without the Fmoc protective group).

Introduction of three levels of molecular diversity (library D)
The levels of molecular diversity were introduced by a similar procedure as described above in the preparation of libraries A, B and C. For the first level of molecular diversity, five stock solutions, each containing 2.5 equivalents (5x) of Fmoc-protected amino acid [Fmoc-L-Phe-OH (496 mg, 1.28 mmol), Fmoc-L-4-Pal-OH (Fmoc-L-4-pyridinealanine-OH) (497 mg, 1.28 mmol), Fmoc-L-Asp(OtBu)-OH (527 mg, 1.28 mmol), Fmoc-L-Ser(Trt)-OH (729 mg, 1.28 mmol) or Fmoc-L-Tyr(2-Cl-Trt)-OH (870 mg, 1.28 mmol)], 2.5 equivalents (5x) of PyBrOP (595 mg, 1.28 mmol) and 2.5 equivalents (5x) of HOBt (175 mg, 1.28 mmol), were prepared in dry DMF (10 mL), and 5 equivalents (5x) of DIPEA (450 µL, 2.56 mmol) were added to each stock solution. To each of the 25 resins 158, was added 2.0 mL of the appropriate stock solution. The mixtures were shaken under an argon atmosphere for 4 h at room temperature, then filtered, washed with DMF (2×), DCM (2×) and MeOH (3×), and dried under a vacuum to afford 5 groups of resins of general structure 159. The completion of the coupling step was verified by acidic mini-cleavage and TLC analysis as described above. The cleavage of the Fmoc group of resins 159 was subsequently performed by a 1 h treatment with 2.0 mL of a solution of piperidine (20%) in DCM. The resins were then filtered, washed with DCM (3×), MeOH (3×) and DCM (1×), and dried under a vacuum to give 5 groups of resins of general structure 160. The second level of molecular diversity was introduced using the same building blocks as for the first level except that Fmoc-L-4-Pal-OH was replaced by Fmoc-L-Ala-OH (398 mg, 1.28 mmol). The procedure described above for the introduction of the first level of molecular diversity was applied without modification. A TLC analysis after acidic mini-cleavage of a sampling of resins 161 confirmed that the introduction of the second level of molecular diversity was completed. Resins of general structure 161 were then treated with 2.0 mL of a solution of piperidine (20%) in DCM for 1 h at room temperature. The resins were filtered, washed with DCM (3×), MeOH (3×)