Discovery of Adamantyl Heterocyclic Ketones as Potent 11β-Hydroxysteroid Dehydrogenase Type 1 Inhibitors

11β-Hydroxysteroid dehydrogenase type 1 (11β-HSD1) plays a key role in converting intracellular cortisone to physiologically active cortisol, which is implicated in the development of several phenotypes of metabolic syndrome. Inhibition of 11β-HSD1 activity with selective inhibitors has beneficial effects on various conditions, including diabetes, dyslipidemia and obesity, and therefore constitutes a promising strategy to discover novel therapies for metabolic and cardiovascular diseases. A series of novel adamantyl heterocyclic ketones provides potent and selective inhibitors of human 11β-HSD1. Lead compounds display low nanomolar inhibition against human and mouse 11β-HSD1 and are selective with no activity against 11β-HSD2 and 17β-HSD1. Selected potent 11β-HSD1 inhibitors show moderate metabolic stability upon incubation with human liver microsomes and weak inhibition of human CYP450 enzymes.


Introduction
11b-hydroxysteroid dehydrogenase isozymes (11b-HSDs) are microsomal enzymes from the short-chain dehydrogenase/reductase family that catalyse the intracellular conversion of physiologically active glucocorticoids and their inert 11-keto counterparts in specific tissues. [1,2] The 11b-hydroxysteroid dehydrogenase type 1 (11b-HSD1), highly expressed in liver, adipose tissue and the central nervous system, acts as an NADPHdependent reductase converting cortisone (1) in humans to the active glucocorticoid cortisol (2) (Scheme 1). The pre-receptor activation of cortisone mediated by 11b-HSD1 provides a mechanism for specific tissues to produce intracellular, nonadrenal cortisol, thereby locally intensifying the glucocorticoid action. [3] Conversely, the 11b-HSD2 isoform is exclusively NAD + dependent and mainly expressed in mineralocorticoid target tissues, such as the kidney and colon. The 11b-HSD2 isoform catalyses the transformation of cortisol to inactive cortisone and reduces the local concentration of cortisol in specific tissues. This mechanism prevents cortisol occupation of mineralocorticoid receptors in the kidney, which may result in sodium retention, hypokalaemia and hypertension. [4][5][6] The tissue-specific regulation of glucocorticoid by 11b-HSDs is of particular physiological importance, as cortisol in humans plays essential roles in the regulation of carbohydrate, lipid and bone metabolism, modulation of inflammatory responses, brain function, and stress. Recent investigations have shown that excessive glucocorticoid action is, in many respects, associated with insulin and leptin resistance, leading to the development of type 2 diabetes, obesity and other metabolic and cardiovascular disorders, the major underlying causes of metabolic syndrome. [7][8][9][10] Glucocorticoid receptor (GR) activation stimulates hepatic glucose production, antagonises insulin secretion from pancreatic b-cells and insulin-mediated glucose uptake in peripheral tissues. [11][12][13][14] GR activation also promotes lipolysis and fatty acid mobilisation. [15] In many aspects, metabolic syndrome shares similar characteristics with symptoms of Cushing's syndrome, a systemic glucocorticoid excess condition, characterised by insulin resistance, high adiposity, dyslipidemia, and hypertension. [16] These metabolic abnormities in Cushing's syndrome can be improved to a certain degree by reducing the excessive glucocorticoid action through surgery or glucocorticoid receptor antagonist treatment. [17][18][19] The similarities observed between phenotypes of Cushing's syndrome and metabolic syndrome suggest the possibility of treating the individual indications of metabolic syndrome by glucocorticoid activity suppression. [20] GR signal-11b-Hydroxysteroid dehydrogenase type 1 (11b-HSD1) plays a key role in converting intracellular cortisone to physiologically active cortisol, which is implicated in the development of several phenotypes of metabolic syndrome. Inhibition of 11b-HSD1 activity with selective inhibitors has beneficial effects on various conditions, including diabetes, dyslipidemia and obesity, and therefore constitutes a promising strategy to discover novel therapies for metabolic and cardiovascular diseases. A series of novel adamantyl heterocyclic ketones provides potent and selective inhibitors of human 11b-HSD1. Lead compounds display low nanomolar inhibition against human and mouse 11b-HSD1 and are selective with no activity against 11b-HSD2 and 17b-HSD1. Selected potent 11b-HSD1 inhibitors show moderate metabolic stability upon incubation with human liver microsomes and weak inhibition of human CYP450 enzymes. ling depends not only on the circulating glucocorticoid levels, but also on the pre-receptor activation of glucocorticoid within cells. Since systemic glucocorticoid levels are generally normal in patients with common forms of obesity or overweight type 2 diabetics, [21] it is speculated that the intracellular glucocorticoid concentration, regulated by 11b-HSDs, is responsible for the metabolic abnormalities. Studies show that 11b-HSD1 expression is elevated in adipose tissue of obese subjects suggesting the possibility of tissue-specific local glucocorticoid excess. [22,23] The correlation between 11b-HSD1 activity, obesity and diabetes has also been validated with genetically modified rodent models. Transgenic mice with overexpression of 11b-HSD1 in fat tissue specifically developed symptoms of insulin-resistant diabetes, hyperlipidemia and visceral obesity. [24,25] In contrast, studies with 11b-HSD1 knock-out mice demonstrated that these animals resist stress-induced hyperglycaemia, diet-induced obesity, and have decreased cholesterol and triglyceride levels. [26] Similarly, overexpression of 11b-HSD2 in transgenic mice also results in increased insulin sensitivity, glucose tolerance, and resistance to body-weight gain on a high-fat diet. [27] 11b-HSD1 activity suppression in animal models with selective inhibitors was found to provide beneficial effects on various indications of metabolic syndrome. [28][29][30] Moreover, clinical studies suggest that inhibition of 11b-HSD1 with carbenoxolone, a nonselective inhibitor, increased hepatic insulin sensitivity and decreased glucose production. [31,32] All this evidence supports the idea of treating type 2 diabetes, obesity and other metabolic abnormalities through selective inhibition of 11b-HSD1 activity with small-molecule inhibitors. [33] This approach has attracted considerable interest in the pharmaceutical industry over the last decade, [34][35][36][37] which has led to the discovery of numerous types of potent, selective 11b-HSD1 inhibitors. [37][38][39][40][41][42] Results from a positive proof-of-concept clinical study of the potent 11b-HSD1 inhibitor INCB013739 developed by Incyte provided substantial evidence that the inhibition of 11b-HSD1 can be a viable treatment of type 2 diabetes. It was reported that INCB013739 treatment of type 2 diabetes mellitus patients who failed on metformin monotherapy could significantly improve hepatic and peripheral insulin sensitivity and reduce haemoglobin A1c and fasting plasma glucose. In patients with hyperlipidaemia or hypertriglyceridaemia, treatment with INCB013739 also lowers triglyceride and cholesterol levels. [43,44] To discover novel, clinically useful inhibitors of 11b-HSD1, it is important to have an array of structural types of inhibitor, as the physicochemical properties of the compounds will determine tissue distribution, hypothalamic-pituitary-adrenal axis effects, and, ultimately, clinical utility. Previously, we reported that some adamantyl ethanone derivatives possess high inhibitory activity on the cellular 11b-HSD1 enzyme. [45] Compounds 3 and 4 exhibit potent inhibition of 11b-HSD1 with IC 50 values of approximately 60 nm. To further improve potency, pharmacokinetic properties and physicochemical properties, we performed optimisation on this series of compound using structure-based design. [46] Compounds containing an adamantyl group linked to a heterocyclic unit through a multiatom linker with a car-bonyl group attached to the adamantane, as illustrated by general structure 5, were synthesised ( Figure 1). These target compounds were screened for their inhibitory activity against human 11b-HSD1 in a HEK293 cell-based assay. Selected potent compounds were tested for inhibitory activity against mouse 11b-HSD1. Their selectivities over 11b-HSD2 and 17b-HSD1 were also examined.
Synthesis of the furanylmethylthio 1-adamantylethanone derivative 18 was performed through a nucleophilic substitution, followed by a coupling reaction of ethanthioate with methyl 5-(chloromethyl)furan-2-carboxylate as illustrated in Scheme 3. Hydrolysis of the methyl ester and subsequent amide formation gave compounds 19 and 20. Their corresponding sulfoxide (21,24) and sulfone derivatives (22, 23 and 25) were synthesised by oxidation with m-CPBA (Scheme 3). No sulfoxide product was obtained from the oxidation of 19 under these conditions.

Structure-activity relationships (SAR)
The target compounds were examined for their inhibition against 11b-HSD1 on an HEK293 cell line transfected with the human HSD11B1 gene. Our previous study found that adamantyl ethanone derivatives exhibited 11b-HSD1 inhibitory activity with lead compounds showing IC 50 values in the range of 50-60 nm. [45] Compounds 3 and 4, with a substituted benzyl or phenyl group, respectively, attached to the adamantyl ethanone through an ether linker, demonstrate high potency, good selectivity over 11b-HSD2, and moderate metabolic stability in human liver microsomes. [45] Therefore, the adamantyl ethanone moiety represents a suitable template for further optimisation.
Replacing the phenyl group of compound 3 with 2-thiophene generated compound 6, which shows moderate activity (IC 50 = 410 nm). Moving the sulfur to the 3-position only increases the activity slightly (7, IC 50 = 280 nm). However, the introduction of a 2thiazole ring results in a tenfold improvement in inhibitory activity, generating the highly potent compound 8 with an IC 50 value of 41 nm. The improvement in activity may be due to the added interaction of the nitrogen atom with the protein in the binding site or the change in molecular physicochemical properties. When a methyl group was added to the 5-position of the thiazole ring, the activity reduces slightly (9, Table 1).
To investigate the possible binding mode of the inhibitor, compound 8 was docked into the published X-ray crystal structure of 11b-HSD1 (PDB: 2ILT [47] ) using the GOLD docking program (version 4.1). The best docking solution predicts that the adamantyl group will pack nicely between the nicotinamide ring of the cofactor and Y183, and the carbonyl is able to stack with the nicotinamide amide. The thiazole ring is predicted to have an edge-on interaction with Y177 (~4 ) (Figure 2).
We also evaluated the effect of replacing the oxygen linker with a sulfur atom in the same core structure (Table 2). Com-  (38)(39)(40)(41)(42). Reagents and conditions: a) 2n NaOH; b) 1. 2 n NaOH; 2. HCl, CH 3 OH; c) 1-adamantyl bromomethyl ketone, Et 3 N. Scheme 5. Synthesis of 1,3,4-thiadiazole derivative (45). Reagents and conditions: a) CS 2 , Et 3 N, DMF; b) 1-adamantyl bromomethyl ketone, Et 3 N (54 % yield). pared with 6, the activity of compound 10 increases by nearly threefold with an IC 50 value of 143 nm, suggesting that the sulfur linker may enable more favourable interactions with the protein. Similarly, compound 11 with the furan-2-yl group is equally potent. However, this effect was not observed in compounds with a thiazolyl group in the aromatic region. Compounds 12 (IC 50 = 244 nm) and 13 (IC 50 = 327 nm) exhibit reduced activity compared to their oxygen linker countparts 8 and 9, respectively, indicating that the linker change may alter the binding position of the thiazolyl ring. Interestingly, we found that the change of the sulfide linker to a sulfoxide greatly raises the inhibitory activity to under 50 nm for both thiophenyl and furanyl derivatives (14,15), suggesting that the oxygen on the sulfur possibly forms further interactions with the enzyme or alters the geometry of the molecule placing the adamantyl and/or the aromatic ring in a position to gain further binding interactions in the active site. Compound 16, with a sulfone linker, exhibits similar potency compared with the sulfide analogue 10, but a fourfold weaker potency compared with the sulfoxide compound 14. However, compound 17 with a furanyl ring maintains the same potency (IC 50 = 36 nm) as its sulfoxide analogue 15. A docking study with 17 found that there is a possible hydrogen-bond interaction from S170 to the carbonyl and from Y183 to either the carbonyl or one of the SO 2 oxygen atoms.
Based on the positive results from the adamantyl ethanone furanyl series, the possibility of altering the substitution on the furan ring was explored (Table 3), as our previous work indicat-    ed that a hydrogen-bonding group on the aromatic ring might potentially improve the inhibition of 11b-HSD1. [45] Surprisingly, the 5-methyl ester derivative 18 shows reduced activity (IC 50 = 264 nm) compared with the unsubstituted compound 11 (IC 50 = 163 nm). Although the sulfoxide linker analogue 21 regains the activity by 2.6-fold, with an IC 50  From the above SAR data, we found that the core structure of adamantyl ethanone linked to methyl 2-thiophenyl or methyl 2-furanyl moieties through a sulfoxide or sulfone group normally generates potent inhibitors of 11b-HSD1. Compounds 14, 15 and 17, with IC 50 values below 50 nm, are the most potent inhibitors from this new series (Table 2 and Table 3).
To investigate the SAR of adamantyl ethanone derivatives with the heterocyclic ring attached directly to a sulfur linker, we synthesised four series of compound containing different 5-membered ring systems and evaluated their inhibition ( Table 4). The thiophenyl derivative 26 (IC 50 = 101 nm) shows activity similar to analogue 10, which has an extra methylene unit between the sulfur and aromatic ring. Replacing the thiophene ring with 1-methyl-1H-imidazole or 5-methyl-1,3,4-thiadiazole gave two compounds with slightly improved potency (27: IC 50 = 93 nm; 29: IC 50 = 61 nm). Interestingly, when a 4methyl-4H-1,2,4-triazole moiety was introduced, the most potent compound 28, with an IC 50 value of 19 nm, was identified. The 4H-1,2,4-triazole core structure can be found in many potent 11b-HSD1 inhibitors reported previously. [48][49][50][51] The dramatic increase in activity could be the result of a network of interactions with the triazole ring. A docking study with compound 28 found that the highest docking score solution poses the triazole ring close to the cofactor with the triazole methyl group fitting in a pocket formed by I121, T124 and Y183. The carbonyl group is predicted to form hydrogen bonds to Y183 and S170 (both~2.5 ). The adamantyl group is predicted to fit tightly in a hydrophobic pocket formed by L171, Y177 and L217 ( Figure 3).
All compounds with a sulfoxide linker (30-33) are highly potent. Among them, the 4H-1,2,4-triazole derivative 32 and the 5-methyl-1,3,4-thiadiazole derivative 33 are the most potent, with IC 50 values of 27 nm and 26 nm, respectively. In comparison with their sulfide linker analogues, compounds 30, 31 and 33 all exhibit nearly double the potency. This observation is in agreement with that seen in the previous series ( Table 2), indicating that the sulfoxide linkers may not only change the geometry of the molecule but may also act as hydrogen-bond acceptors with amino acid residues in the active site of 11b-HSD1.
The sulfone version of this series displays mixed results. The thiophenyl derivative 34, the least potent in this series, loses 11-fold activity compared with the sulfoxide analogue 30. Similarly, triazole derivative 36 (IC 50 = 95 nm) has an activity 3.5-fold lower than the sulfoxide analogue 32, and fivefold lower than the sulfide analogue 28. However, compound 37 (IC 50 = 31 nm) still retains the same level of potency as the compound with a sulfoxide linker (33, Table 4). As one of the most potent compounds with a sulfone linker, this compound is of particular interest for further investigation.
Having found that the 4H-1,2,4-triazole and 5-methyl-1,3,4thiadiazole derivatives display significant inhibition against 11b-HSD1, we investigated the SAR for variation of substituents with different size and electronic properties on the aromatic ring (Tables 5 and 6). The 5-methyl substituent on the 4H-1,2,4-triazole ring gives a compound with relatively high potency (38, IC 50 = 36 nm; Table 5); however, it is approximate- ly twofold less active than the unsubstituted derivative 28. Substitution with a hydrophobic cyclopropyl group (40, IC 50 = 129 nm) results in a further decrease in activity by a further 3.6-fold, suggesting that the size of the substituent permitted in the aromatic region could be limited. The 5-thiophenyl-substituted compound 44 exhibits very low activity with an IC 50 value of 5871 nm. This dramatic loss in activity may be due to an alteration of the binding mode caused by the biaryl unit or the physiochemical properties of the molecule. The introduction of a group with hydrogen bonding capacity at the 5-position keeps the IC 50 value in the range of 100-150 nm (41 and 42, Table 5); however, these compounds are three-to fourfold less active than the 5-methyl analogue 38 and much less active than the unsubstituted compound 28. Replacement of the methyl group of compound 38 with an iso-propyl group leads to a 3.8-fold loss of activity (39, IC 50 = 137 nm), suggesting that the bulkier groups are not favoured in that region. This observation is in agreement with the result for compound 43 (IC 50 = 1857 nm), with cyclopropyl substituents at both the 1-and 5-positions.
Compound 46, with an IC 50 value of 306 nm, shows 20 % of the activity of the methyl-substituted 29. The 5-N,N-dimethylamino substituent also has a strongly negative effect on the activity (45, IC 50 = 4005 nm), suggesting the possibility of limited steric and/or electronic requirements in a confined region.
It is interesting to note that the amino-substituted compound (48), with an IC 50 value of 71 nm, shows nearly the same level of potency as the methyl-substituted compound 29. The benefit of having an amino group on the 1,3,4-thiadiazole ring is that it can provide a handle for further optimisation. A docking study suggested the possibility of the thiadiazole amino group of 48 forming hydrogen bonds with NADP phosphate (3.4 ), and the carbonyl group interacting with Y183 (2.8 ) and S170 (2.5 ) also through hydrogen bonds. It was surprising to see the sulfone-linked analogue 49 exhibiting only modest activity with an IC 50 value of 1787 nm. The acetamide substituent (50, IC 50 = 848 nm) reduces the activity of 48 by almost 12-fold, providing further proof that the size of the substituent is crucial to the activity. Although oxidation of the sulfide linker of 50 to sulfoxide or sulfone raised the inhibitory activity by about 2.5-fold, compounds 51 (IC 50 = 333 nm) and 52 (IC 50 = 309 nm) are still far less potent than their methylsubstituted analogues 33 and 37.
Having identified ten potent compounds, with IC 50 values below 50 nm, from these adamantyl ethanone derivatives, we examined their inhibition against mouse 11b-HSD1, since a mouse model will most likely be used in vivo. Compounds 8, 14, 15, 33 and 37 all exhibit similar levels of potency against the mouse and human enzymes, which makes them suitable for further evaluation in mouse models. Although compounds 17, 28, 31 and 38 are several fold less active on the mouse enzyme, they still show relatively strong activity, especially    Table 7). Compounds 17, 28, 31 and 38 were also tested for their inhibition against human 11b-HSD2 and 17b-HSD1. At a concentration of 10 mm, these compounds are all inactive (data not shown) and therefore are regarded as highly selective 11b-HSD1 inhibitors.
To improve the solubility of an adamantyl ethanone derivative, compound 37 was transformed into the enolate sodium salt (53) (Scheme 6), which exhibits enhanced water solubility (> 4 mgmL À1 ) in a Na 2 HPO 4 buffer solution with or without 5 % N,N-dimethylacetamide (DMA) as a cosolvent. Compound 53 displays relatively high potency against 11b-HSD1 with an IC 50 value of 108 nm.

Permeability, metabolic stability and CYP450 inhibition studies
The membrane permeability of two selected compounds (14 and 31) was determined using the Caco-2 cell model. The apparent permeability coefficient (P app ) of each drug from the apical to basolateral side was measured at a concentration of 20 mm. The results indicate that both compounds possess high permeability (14: P app (A > B) = 2.2 10 À5 ; 31: P app (A > B) = 4.4 10 À5 ), which supports the fact that these molecules are of a lipophilic nature.
Potent inhibitors 15, 17, 28 and 37 were evaluated for their metabolic stability in human liver microsomes (Table 8). Compounds 15 and 37 both suffer from rapid metabolism under the test conditions, with a half-life of only 30-40 min and a high clearance rate. However, triazole derivative 28 shows im-proved metabolic stability with a half-life of more than 1 h and an intrinsic clearance of~9 mL min À1 mg À1 . More interestingly, compound 17 is relatively stable under the human liver microsomes incubation conditions, displaying a half-life of more than 4 h and a clearance rate below 3 mL min À1 mg À1 . This clearly indicates that metabolic stability suitable for a compound to be dosed in vivo may be achieved with this structural template.
To examine the possibility of these compounds interfering with the metabolism of other drugs, compounds 14, 17, 28 and 37 were tested for their inhibition of key human cytochrome P450 enzymes: 1A2, 2C9, 2C19, 2D6, 3A4-BFC and 3A4-BQ (Table 9)

Conclusions
In summary, we have identified several new series of adamantyl ethanone heterocyclic derivatives as novel inhibitors of human and mouse 11b-HSD1. The ten most potent compounds (Table 7) display selective high inhibitory activity  [a] The parent compound is incubated at 37 8C with human liver microsomes in presence of the cofactor NADPH for 40 min. Disappearance of the parent compound is monitored using a HPLC system. Half-life (t 1/2 ) and intrinsic clearance (CL int ) values were calculated accordingly. [52] Data reported as the mean value of two experiments. against 11b-HSD1, with IC 50 values below 50 nm, when evaluated on a stably transfected HEK293 cell line. It was found that an adamantyl ethanone motif tethered to a 5-membered aromatic heterocycle through an oxygen, sulfur, sulfoxide or sulfone linker provides a suitable pharmacophore for inhibition of 11b-HSD1. The further optimisation of the previously reported compounds 3 and 4 lead to a potent inhibitor with approximately a threefold improvement in activity. Compound 28 (11b-HSD1, human IC 50 = 19 nm, mouse IC 50 = 61 nm) displays reasonable metabolic stability under human liver microsomes incubation conditions with a half-life of 68 min and very weak inhibition of key human CYP450 enzymes. This compound, in particular, is regarded as a candidate for further preclinical investigation.
Experimental Section 11b-HSD1 inhibition cellular assay using a scintillation proximity assay (SPA) protocol: [53] Wild type HEK293 cells lack endogenous 11b-HSD1 activity and this cell line has been shown to be a suitable system for evaluating 11b-HSD1 activity after being transfected with the plasmid for expression of 11b-HSD1 or 11b-HSD2. The enzyme activity was determined by measuring the amount of tritiated product by using a scintillation proximity assay (SPA). The high-throughput cell-based assays were conducted on the HEK293 cell line stably transfected with the HSD11B1 gene using modified literature protocols. Cells were incubated in 96-well micro-plates in the presence of tritiated substrate and the assay plates contained internal high and low controls to allow calculation of percentage inhibition. Each well of a 96-well culture plate was seeded with HEK293/HSD11B1 cells in 100 mL medium. When the cells were 80 % confluent, the medium was removed from each well then 100 mL of fresh, serum-free, medium containing [ 3 H]cortisone and test compound in 1 % DMSO was added to each well. The control wells were also dispensed. The high control wells did not contain compound, while low controls did not contain cells. The plate was incubated at 37 8C for the required time period, after which, 50 mL of media was removed from each well and transferred to a microplate containing 100 mL of a pre-incubated mixture of anti-cortisol antibody and SPA bead. The mixture was incubated with gentle shaking until equilibrium was reached, before transferring to a scintillation counter to establish the enzyme activity in each sample.
Docking study procedure: Selected ligands were docked into the human 11b-HSD1 protein X-ray crystal structure (PDB: 2ILT [47] ) using the GOLD docking program (version 4.1) with default settings in the presence of the cofactor. The binding site was defined as a sphere of 10 radius around the centroid of the ligand in the 2ILT structure. Each ligand was docked 25 times. The GOLDscore scoring function was used to rank the ligands in order of fitness to generate the best poses of the ligand in the active site.
General methods for synthesis: 1-Adamantyl bromomethyl ketone was obtained from Aldrich Chemical Co. (Gillingham, UK). All other chemicals were purchased from either Aldrich Chemical Co. or Alfa Aesar (Heysham, UK). All organic solvents of AR grade were supplied by Fisher Scientific (Loughborough, UK). Melting points were determined using a Stanford Research Systems Optimelt MPA100 and are uncorrected. Compounds in solid form were crystallised from CH 2 Cl 2 /EtOAc. Thin layer chromatography (TLC) was performed on pre-coated aluminium plates (Merck, silica gel 60 F 254 ). Products were visualised by UV irradiation at 254 nm and staining with 5 % w/v molybdophosphoric acid in EtOH, followed by heating. Flash column chromatography was performed on prepacked columns (RediSep R f ) and gradient elution (solvents indicated in text) on the Combiflash RF system (Teledyne Isco). 1   Method C: Synthesis of the adamantyl ethanone sulfoxide and sulfone derivatives 14-17: A cold solution of the corresponding sulfanyl derivative (1 equiv) in CH 2 Cl 2 (10 mL) was treated with m-CPBA (2 equiv), and the mixture was stirred at À10 8C to 08C for 40 min. The mixture was then partitioned between CH 2 Cl 2 and 5 % aq Na 2 CO 3 , and the organic phase was washed with brine, dried over MgSO 4 , filtered and concentrated in vacuo to give the crude product. The sulfoxide and sulfone were separated using flash chromatography (EtOAc/CH 2 Cl 2 ; gradient elution).