Direct Stereoselective Aziridination of Cyclohexenols with 3‐Amino‐2‐(trifluoromethyl)quinazolin‐4(3H)‐one in the Synthesis of Cyclitol Aziridine Glycosidase Inhibitors

Cyclophellitol aziridine and its configurational and functional isomers are powerful covalent inhibitors of retaining glycosidases, and find application in fundamental studies on glycosidases, amongst others in relation to inherited lysosomal storage disorders caused by glycosidase malfunctioning. Few direct and stereoselective aziridination methodologies are known for the synthesis of cyclophellitol aziridines. Herein, we present our studies on the scope of direct 3‐amino‐2‐(trifluoromethyl)quinazolin‐4(3H)‐one‐mediated aziridination on a variety of configurational and functional cyclohexenol isosters. We demonstrate that the aziridination can be directed by an allylic or homoallylic hydroxyl through H‐bonding and that steric hindrance plays a key role in the diastereoselectivity of the reaction.


Introduction
Glycosidases are enzymes involved in the degradation of complex glycoconjugates in nature and are of relevance both in biomedicine and biotechnology. [1] Many glycosidases follow a two-step Koshland double displacement mechanism, which involves a covalent enzyme-glycoside intermediate. [2] The active site of such retaining glycosidases is usually composed of an aspartic acid or glutamic acid, termed the catalytic acid/base, and an aspartate/glutamate (or occasionally a tyrosine) termed the nucleophile. In the first step of substrate hydrolysis, the exocyclic oxygen is protonated by the acid/base residue. Next, the catalytic nucleophile attacks at the anomeric carbon and effects an S N 2 displacement of the aglycon, yielding a covalent enzyme-glycoside complex with inversion of the anomeric stereochemistry.
In the second step, a water molecule is deprotonated by the acid/base carboxylate and hydrolyses the enzyme-substrate intermediate with a second inversion of the anomeric configuration ( Figure 1A). [3] Cyclitol aziridines can mimic the conformation of the oxocarbenium ion transition state and irreversibly 1 present our studies on the scope of direct 3-amino-2-(trifluoromethyl)quinazolin-4(3H)-one-mediated aziridination on a variety of configurational and functional cyclohexenol isosters. We demonstrate that the aziridination can be directed by an allylic or homoallylic hydroxyl through H-bonding and that steric hindrance plays a key role in the diastereoselectivity of the reaction. inactivate glycosidases by covalently reacting with the nucleophilic carboxylate ( Figure 1B). Based on this virtue and the fact, as supported by several studies in recent years from our group, [4] that covalent and irreversible inhibition is often both very effective and highly selective, cyclitol aziridines and their corresponding activity-based probes (ABPs) are highly useful tools for chemical glycobiology research. [5] An important step in the synthesis of cyclitol aziridine inhibitors and ABPs involves the stereoselective aziridination of a suitable cyclohexene precursor. [6] In contrast to epoxidations, synthetic methodologies for the direct and stereoselective aziridination of alkenes are scarce. [7] Our previous work on the synthesis of cyclophellitol aziridines relied mostly on either intramolecular iodocyclization followed by aziridine formation (Figure 2A) [5a,c,d] or Staudinger-type ring closure of 1,2-azidoalcohols obtained from an epoxide precursor ( Figure 2B). [5b] We also investigated the use of O- (2,4-dinitrophenyl)hydroxylamine (DPH) as nitrogen donor with Rh 2 (esp) 2 as catalyst ( Figure 2C). [8] However, some limitations were encountered with these methodologies. Although the iodocyclization/intramolecular substitution sequence gives complete stereochemical control with reasonable to good overall yields, this sequence is not an op- Eur. J. Org. Chem. 0000, 0-0 www.eurjoc.org tion when the desired aziridine has the opposite stereochemistry of that of the directing alcohol moiety. The rhodium-catalyzed aziridination methodology appeared to be non-stereoselective and to proceed in relatively low yields, [8] while in the Staudinger approach, purification can be challenging and overreduced amino-alcohols are occasionally generated as side products (although the use of polymer-bound triphenylphosphine alleviates these shortcomings to certain extent [8b,9] ). In the late 1960s, Atkinson and co-workers described the lead acetate mediated in situ oxidation of diverse hydrazines, amongst which 3-aminobenzoxazolin-2-one was used as aziridinating agent. [10] They initially proposed that nitrenes could be formed as intermediates [10,11] which would react with electrophilic and nucleophilic olefins to give the observed aziridines. Atkinson et al. extensively re-examined the reactivity of different substituted 3-aminoquinazolin-4(3H)-ones (Q-NH 2 s) towards diverse olefins, [12] and concluded that N-acetoxyaminoquinazolones (Q-NHOAc) are the reactive intermediates, rather than nitrenes. [13] Recently, the stereospecific aziridination of a partially protected galacto-configured cyclohexene has been described by Llebaria and co-workers based on the use of 3-amino-2-ethylquinazolin-4(3H)-one (Et-Q-NH 2 ) and PhI(OAc) 2 (PIDA). The thus formed -galacto-configured acetylated aziridine was employed in this study to produce N-aminoaziridine based irreversible inhibitors ( Figure 2D). [14] Inspired by this work of Llebaria and co-workers we decided to explore the scope of the direct olefin aziridination reaction by investigating the reactivity of diverse aminoquinalozolin-4-ones towards differently configured and functionalized cyclohexenol substrates ( Figure 2E). As we show here, this reaction proved to be particularly well suited for the synthesis of α-L-idose-configured cyclophellitol aziridine, a key intermediate for the synthesis of new α-L-iduronidase inhibitors and ABPs, as we recently reported in a separate body of work. [15]

Results and Discussion
As the first research objective, D-gluco-configured cyclohexene 1a was used as starting material to screen the most promising aminoquinazolinones described as nitrogen donors in the literature: 3-amino-2-ethylquinazolin-4(3H)-one (Et-Q-NH 2 ), 3amino-2-(trifluoromethyl)quinazolin-4(3H)-one (CF 3 -Q-NH 2 ) and the chiral (S)-3-amino-2-(1-hydroxy-2,2-dimethylpropyl)quinazolin-4(3H)-one (HO-Q-NH 2 ), which all form in situ the reactive Nacetoxy-aminoquinazolinones in the presence of PIDA. In line with previous results, [12e] CF 3 -Q-NH 2 gave superior yields (69-75 % of 1b) when using cyclohexene 1a, PIDA and the quinazolone in a 1:2:2 ratio respectively and by forming the reactive Nacetoxy-aminoquinazolinone at -78°C prior to addition of the olefin at -23°C. Aziridine intermediate 1d was isolated in 54 % yield when using HO-Q-NH 2 as aziridination agent, whereas reaction with Et-Q-NH 2 returned starting material only (Scheme 1). Notably, the -gluco-configured aziridine was formed stereoselectively in a 1.5 mmol reaction scale, indicating that hydrogen bonding from the homoallylic alcohol C7-OH guides the incoming Q-NHOAc, in agreement with the mechanistic proposal of Atkinson et al. (Scheme 2). [12a] These results led us to further investigate aziridinations with CF 3 -Q-NH 2 on different cyclohexene substrates. 2b was exclusively formed in 55 % yield, providing further support for H-bonding guided delivery of the aziridinating reagent (Scheme 2). When perbenzylated gluco-cyclohexene 3a was subjected to the same reaction conditions no conversion was observed, indicating that the system is not reactive enough without the hydrogen bonding guided delivery and/or that the double bond, with relatively bulky substituents on either side of the alkene, is too crowded to allow for an effective addition. A similar pattern was observed with galacto-configured cyclohexenes. Galacto-configured cyclohexene 4a could be stereoselectively transformed into -aziridine 4b while cyclohexene 5a afforded α-aziridine 5b in 61 % yield (Scheme 3). These examples again illustrate the impact of neighboring (homo)allylic alcohol functionalities. Partially protected conduritol 6a was also amenable to stereoselective aziridination, affording -aziridine 6b in 66 % yield, whereas the starting material was recovered when the reaction was performed with perbenzylated conduritol 7a (Scheme 4). In order to investigate whether an alcohol further away from the alkene could guide the reagent to one of the diastereotopic faces of the double bond, we examined the aziridination of partially protected xylo-configured cyclohexene 8a. In this case only -isomer 8b was obtained, indicating that the 4-OH is too distal for a productive H-bond interaction and that the aziridination takes place on the least hindered face of the double bond, opposite of the C-2-benzyl ether (Scheme 5). We finally explored aziridination of L-ido-configured cyclohexenes in order to obtain α-L-ido-configured aziridines as potential intermediates for the development of new iduronidase inhibitors. [15] Partially protected cyclohexene 9a was not amenable to aziridination, possibly because the primary alcohol directs to the beta side while this region may be hindered by the allylic benzyl ether. Considering that -D-xylo-configured aziridine 8a was obtained without H-bonding, we postulated H-bonding would not be essential for a satisfactory aziridination in case the double bond is readily accessible. To test this hypothesis, the free alcohols in 9a were benzylated (benzyl bromide, sodium hydride) to generate cyclohexene 10a. From this fully protected cyclohexenol, α-L-aziridine 10b was obtained in 43 % yield together with 32 % recovered starting material after direct aziridination using CF 3 -Q-NHOAc as the nitrogen donor (Table 1, Scheme 6). The α-configuration of aziridine 10c was confirmed by comparison of the experimental 1 H NMR coupling constants with the corresponding calculated values acquired from DFT calculations. [15]  In all cases, one step deprotection of the aziridine and hydroxyls in the aforementioned CF 3 -Q functionalized aziridine intermediates was achieved under Birch conditions using lithium and liquid ammonia at -78°C. The reactions were quenched with H 2 O and impurities derived from CF 3 -Q precipitated and were removed by filtration. The cyclitol aziridines were finally obtained in excellent yields after cation-exchange chromatography with Amberlite H + resin to eliminate the lithium hydroxide salts (Table 1, 81-99 %).

Conclusions
We have explored direct aziridination of both, partially protected and fully protected, configurational cyclohexenol using different substituted 3-aminoquinazolin-4(3H)-ones. From these studies we identified 3-amino-2-(trifluoromethyl)quinazolin-4(3H)-one (CF 3 -Q) as the superior aziridinating agent. Using this reagent, direct aziridination reaction can be applied on diverse glycoside configured cyclohexenes, and it appears that aziridination can be directed by allylic or homoallylic hydroxyls through H-bonding and that steric hindrance plays an essential role in the diastereoselectivity of the reaction. With this in mind, one could tune the cyclohexene scaffold depending on the desired configuration of the target aziridine and thus, synthesize diverse glycosidase inhibitors effectively in asymmetric fashion.
Note: Numbering of proton peaks in cyclohexene and cyclitol aziridine derivatives is according to the numbering in Scheme 1.