Stereo- and Enantioselective Addition of Organolithiums to 2-Oxazolinylazetidines as a Synthetic Route to 2-Acylazetidines

A new synthetic route to N-alkyl-2-acylazetidines was developed through a highly stereoselective addition of organolithiums to N-alkyl-2-oxazolinylazetidines followed by acidic hydrolysis of the resulting oxazolidine intermediates. This study revealed an unusual reactivity of the C=N bond of the oxazoline group when reacted with organolithiums in a non-polar solvent such as toluene. The observed reactivity has been explained considering the role of the nitrogen lone pair of the azetidine ring as well as of the oxazolinyl group in promoting a complexation of the organolithium, thus ending up with the addition to the C=N double bond. The high level of stereoselectivity in this addition is supported by DFT calculations and NMR investigations, and a model is proposed for the formation of the oxazolidine intermediates, that have been isolated and fully characterized. Upon acidic conditions, the oxazolidine moieties were readily converted into 2-acylazetidines. This synthetic approach has been applied for the preparation of highly enantioenriched 2-acylazetidines starting from chiral not racemic N-alkyl-2-oxazolinylazetidines.


INTRODUCTION
The four-membered saturated heterocycle azetidine is a valuable scaffold exploited in several active research areas (Singh et al., 2008;Couty and Evano, 2009;Antermite et al., 2017). The strain and dynamic phenomena associated to the azetidine ring allows exploring new chemical space for organic synthesis and drug discovery purposes (Degennaro et al., 2014a,b;De Ceglie et al., 2011). Recently, different approaches have been reported for the synthesis of azetidine derivatives for targeting lead compounds and bioisosteres for drug discovery (Couty et al., 2004;Ferraris et al., 2007;Pérez-Faginas et al., 2011). Despite the great interest for this small-sized heterocycle, some structural motifs, such as 2-ketoazetidines, seems to be poorly investigated. Biologically active compounds, incorporating the 2-acylazetidine moiety, include natural products such as alkaloids found in the genus Daphniphyllum (Kobayashi and Kubota, 2009) as well as 2-aroylazetidines known as potent inhibitors of dipeptidyl peptidase IV (DPP-IV) (Ferraris et al., 2004), a prolinespecific serine protease used as target in several therapeutic areas such as diabetes (Weber, 2004) pain (Ronai et al., 1999), and cognition enhancement (During et al., 2003; Figure 1). FIGURE 1 | Examples of 2-acylazetidine motif in natural products and biologically active compounds. SCHEME 1 | Reported synthesis of 2-acylazetidines.
In 2011, our group reported an intriguing approach for the synthesis of 2-ketoaziridines exploiting an unprecedented reactivity profile of chiral 2-oxazolinylaziridines when subjected to reaction with organolithiums (Degennaro et al., 2011). This work focused on a regioselective α-lithiation of optically active N-phenylethyl-2-oxazolinylaziridines by using a strong base such as n-buthyllithium (n-BuLi) in a coordinating solvent (THF) at −78 • C within 1 h. SCHEME 2 | Example of reactivity controlled by dynamics in small nitrogenated heterocycles.
The resulting lithiated species proved to be chemically and configurationally stables, under the experimental conditions, and were trapped with electrophiles in highly stereoselective manner (Scheme 2, path A). In particular, this study demonstrated the presence of a mixture of two equilibrating invertomers for N-alkyl-2-oxazolinylaziridines as a result of the nitrogen inversion (Capriati et al., 2002;Luisi et al., 2005; Scheme 2). It was demonstrated by NMR and DFT calculations, that the dynamics at nitrogen was dependent on the nature of the substituent at the C2 of the azetidine ring. A dynamic model was proposed to account for a diverse and unexpected reactivity observed in a nonpolar solvent such as toluene. In fact, as a consequence of a competing complexation of the organolithium reagent with the lone pair of the aziridine nitrogen led to an unusual attack to the C=N bond of the oxazoline ring with consequent formation of an oxazolidine intermediate, useful precursor of the corresponding 2-acylaziridine (Scheme 2, path B). Better results in term of yield and stereoselectivity were obtained by reacting C2-substituted oxazolinylaziridines with organolithiums in toluene and at higher temperature (0 • C). Therefore, this model suggests that controlling the nitrogen dynamics, in these systems, could be possible to address the reactivity providing both α-substituted aziridines and 2-ketoaziridines.
Inspired by these preliminary results on 2oxazolinylaziridines, highlighting the role of nitrogen dynamics on the reactivity of this three-membered heterocycles, we were keen to build on this ground a stereoselective synthesis of 2-ketoazetidines, likely starting from the corresponding 2-oxazolinyl azetidines (Scheme 2). We are glad to report herein the results obtained in this investigation.

Substrates Synthesis
The first step for the synthesis of 2-oxazolinylazetidines 4a (R, R)-4b and (R, S)-4b, reported in Scheme 3, involved the reaction of methyl 2,4-dibromobutanoate 1, commercially available, with a suitable primary amine. When ester 1 was reacted with benzylamine, azetidinylester 2a was obtained in high yield, and, by subsequent treatment with 2-amino-2-methylpropanol, nhexyllithium, in presence of catalytic amount of LaCl 3 afforded hydroxamide 3a (Jang et al., 2015; Scheme 3). The last step consisted in an intramolecular cyclization of hydroxamide 3a mediated by diethylaminosulfur trifluoride at low temperature, that provided the desired oxazoline 4a. A similar protocol was applied to the preparation of 2-methylcarboxylates (R, R)-2b, and (R, S)-2b, by using (R)-1-phenylethylamine as chiral not racemic amine, to yield orthogonally protected and chromatographically separable diastereomeric mixture of esters in 64% yield in 1:1 ratio. Hydroxamides (R, R)-3b and (R, S)-3b where separately synthesized as previously described for amide 3a and, in the case of amide (S, S)-3b, prepared from (S)-1-phenylethylamine, the structure has been solved by X-Ray analysis, confirming the chemical structure and the absolute stereochemistry of the compound (see Supplementary Material). Finally, the cyclization step yielded 2-oxazolinylazetidines (R, R)-4b and (R, S)-4b, as single diastereoisomers, each with excellent enantiomeric ratio (er = 98:2). SCHEME 3 | Substrates synthesis.
providing the corresponding tertiary alcohol as the main product (Couty et al., 2011). This prompted us to consider the approach reported in Scheme 2 as an alternative route to 2-acylazetidines. We started our investigation considering the reaction of azetidine 4a with n-BuLi in toluene (Table 1). Pleasingly, oxazolidine 5a was the sole product observed in these experiments with no evidence for the α-deprotonation of the azetidine ring.
By adding n-BuLi at room temperature, within 20 min, azetidine 5a was obtained in 30% yield, after aqueous work up, as a 82:18 mixture of diastereomers ( Table 1, No. 1). With the aim to improve both yield and diastereoselectivity, the reaction was conducted at lower temperature. To our delight, running the reaction at −78 • C furnished the product 5a in high yield and as single diastereoisomer ( Table 1, No. 4). The use of an excess n-BuLi, considerably lowered the yields, giving also a complex mixture of products ( Table 1, No. 5-6). We considered the conditions in No. 4 as optimal for the examination of the reaction scope. Under optimized conditions, the nucleophilic addition of n-hexylLi, MeLi and EtLi proceeded in highly stereoselective manner affording compounds 5b-d as single diastereoisomer with satisfactory yields (Scheme 4). By using PhLi as nucleophile, the reaction required longer times (3 h) and product 5e was obtained with 82% yield and 98:2 dr. Surprisingly, the reaction with 2-thienyllithium did not afford any product, even keeping the reaction up to 6 h at −78 and 0 • C, presumably as a consequence of the presence of ethereal solvents, namely THF, in 2-thienyllithium solution. Based on these results, the same protocol (i.e., toluene, −78 • C), was tested using optically active oxazolinylazetidine (R, R)-4b.
stereogenic center of the oxazolidine ring. All attempts to get suitable crystals for X-ray analysis were unsuccessful, and NOESY experiments were inconclusive. Based on our previous experience on the stereochemical assignments in three and four-membered heterocycles by using DFT calculations and NMR predictions, we decided to pursue this approach for solving this stereochemical puzzle (Azzena et al., 2018). The diastereoiomers (R, R, R)-6c and (R, R, S)-6c were considered. A relaxed potential energy surface (PES) scan of the dihedral angles C 43 N 16 C 17 C 19 and O 2 C 3 C 11 C 12 , which define the relative position of the two substituents (oxazolidine ring and benzyl group) on the azetidinyl ring, was performed. All other parameters were allowed to vary freely during the PES scan; the single point energies were calculated at DFT/B3LYP/3-21G level in vacuo. After conformational minimization, the lowest energy conformers of the two diastereoisomers was subjected to fully unconstrained geometry optimization at SMD/DFT/B3LYP/6-311++G(d,p) level, followed by vibrational analysis. The free energy values provided by the vibrational analysis calculations indicated that diastereoisomer (R, R, R)-6c was 3.8 Kcal/mol more stable with respect to (R, R, S)-6c. At the end, optimized structures were used for prediction of nuclear shieldings using the gauge independent atomic orbital (GIAO) approach. GIAO NMR calculations were performed at SMD(CHCl 3 )/DFT/ MPW1PW91/gen level and the shielding tensors (σ calc ) scaled to obtain the predicted chemical shifts (δscal, see Supplementary Material). The NMR chemical shifts (δ) were calculated as the differences of isotropic shielding constants (σ) with respect to the TMS (tetramethylsilane) reference, calculated at the same level of theory. The pcS-2 basis set, specifically designed for NMR shielding constant calculations (Jensen, 2008), was used for H and C atoms and 6-311++G(d,p) for N and O atoms. For indirect spin-spin coupling constants J HH calculation, we selected the SMD/DFT/B3LYP/6-311++G(d,p) level of theory, a good compromise between accuracy and computational cost. Indeed in our earlier studies, we noticed the performance of this method in predicting proton and carbon NMR shieldings as well as the spin-spin coupling constants J HH (Azzena et al., 2011;Carroccia et al., 2014;Zenzola et al., 2014;Degennaro et al., 2015a,b;Pisano et al., 2016) . All calculations were performed with the Gaussian 09 program at DFT level and the solvent effect was modeled using the self-consistent reaction field (SCRF) calculations within the SMD model (Marenich et al., 2009; Supplementary Material). The statistical parameters CMAE and R 2 of δ scaled /δ expt were determined to establish the consistency between the theoretical and experimental magnetic FIGURE 2 | (A) Calculated spectrum of (R,R,R)-6c; (B) Real spectrum of (R,R,R)-6c; (C) Calculated spectrum of (R,R,S)-6c.
parameters of the two possible diastereoisomers, and the best fit in all cases was found for the diastereoisomer (R, R, R)-6c (see Supplementary Material). In Figure 2 the comparison between experimental and calculated 1 H NMR spectra of azetidine 6c is reported. A better match can be assessed between the real and calculated spectra of (R,R,R)-6c (Figures 2A,B).
Based on the results obtained by DFT calculations, we assumed that the configuration at the new created stereocentre of 6c might be (R). With the aim to rationalize the stereochemical outcome of the nucleophilic addition of organolithiums to the C=N double bond of the oxazoline ring, we considered the stereodynamic model proposed for oxazolinylaziridines (Scheme 2). Assuming that oxazolinylazetidines could show dynamic phenomena associated with both nitrogen inversion and ring puckering (Parisi et al., 2016), the stereochemistry of substrates (R, R)-4b and (R, S)-4b was first assessed by NOESY experiments (see Supplementary Material). This stereochemical evaluation demonstrated a trans relationship SCHEME 6 | Synthesis of 2-acylazetidines.
between the nitrogen substituent and the oxazoline ring. Considering such stereochemical arrangement, the two lone pairs belonging to both azetidine and oxazoline nitrogens could likely be oriented in such a way to promote an easy formation of a complex with the organolithium reagent (Scheme 5). Under these stereochemical restrictions, it is reasonable to foresee the nucleophilic addition of the organolithium to the most accessible face of the planar C=N bond, resulting an (R) configuration at the new stereocenter starting from (R, R)-4b, and (S) configuration starting from (R, S)-4b (Scheme 5). On the basis of this model, the absolute configuration for compounds 6a-d and 7a-b have been supposed to be (R,R,R) and (R,S,S), respectively. Similar reasoning can be made for azetidine 4a leading to adducts 5a-e with (R, * R * ) relative configuration. It is worth pointing out that, in striking contrast to what previously observed for aziridines, where a preferential α-lithiation took place under similar reaction conditions, with the nucleophilic addition occuring in 15-20% extent and low stereoselectivity, in the case of oxazolinylazetidines α-lithiation was never observed even at higher temperature. It is reasonable to assume that such peculiar stereoelectronic requirements realized with azetidines prevented the deprotonation event, leading to a stable complex prone to undergo exclusively nucleophilic addition to the C=N double bond.

Synthesis of 2-acylazetidines
All the products 5a-e were isolated by flash chromatography, even though a partial and expected hydrolysis of the oxazolidine moiety in acidic media by silica gel occurred. This evidence prompted us to explore a mild and easy acidic hydrolysis by using silica gel in dichloromethane. Treatment of 5a-d with SiO 2 in dichloromethane (DCM) as solvent for 3 h afforded Nalkyl-2-acylazetidines 8a-d in quantitative yields (Scheme 6). The oxazolidine 5e resulted unreactive under these conditions, and all the attempts to hydrolyze the cyclic aminal failed. Similarly, the hydrolytic protocol was applied to compounds 6a-c and 7b giving good yields of 2-acylazetidines 9a-c and 10b, respectively, in enantiopure form as confirmed by HPLC analysis (see Supplementary Material). A partial epimerization for azetidine 9c in acidic media by silica gel has been observed, leading to the formation of corresponding diastereoisomer 10c (dr = 50/50). Moreover, complete hydrolysis of oxazolidines 7a and 7c taken place already during purification by chromatography on silica gel giving satisfactory yields of 2-acylazetidines 10a and 10c. The absolute stereochemistry of all optically active compounds was assigned according to data reported in the literature (Couty et al., 2011).

CONCLUSIONS
We have demonstrated that in an apolar solvent, such as toluene, different organolithiums were capable to give an unexpected regio-and stereoselective addition at the C=N bond of the oxazoline group of N-alkyloxazolinylazetidines. Different 1,3oxazolidinyl azetidines formed in high yield and resulted useful precursors of 2-acylazetidines by acidic hydrolysis. The expected deprotonation event in α position with respect to oxazoline ring did not take place in this conditions. With the aim to rationalize the mechanism and the stereochemical outcome of the addition reaction, a stereodynamic model has been proposed, taking into consideration complexation and dynamic phenomena associated with the azetidine's nitrogen inversion. The configuration assignment, performed on oxazolidinyl azetidines as intermediates by NMR and DFT calculations, resulted mandatory for the validation of proposed model. Even though the stereochemical information generated in the addition reaction is lost in the hydrolysis of oxazolidine ring, this work furnishes an outstanding example of reactivity controlled by dynamics of small nitrogenated heterocycles. Work is in progress to further explore the reactivity of 2-oxazolinylazetidines with organolithium in coordinating solvents.

General Information
Flash chromatography was performed using 70-230 mesh Al 2 O 3 (either neutral or basic activity II-IV), with the indicated solvent system according to standard techniques. Analytical thin layer chromatography (TLC) was carried out on precoated 0.25 mm thick plates of Kieselgel 60 F254; visualization was accomplished by UV light (254 nm) or by spraying a solution of 5 % (w/v) ammonium molybdate and 0.2 % (w/v) cerium(III) sulfate in 100 ml 17.6 % (w/v) aq. sulphuric acid and heating to 200 • C for some time until blue spots appear. Infrared spectra (v max , FT-IR) were recorded in reciprocal centimeters (cm −1 ). Nuclear magnetic resonance spectra were recorded on 300 or 500 MHz spectrometers. The frequency used to record the NMR spectra is given in each assignment and spectrum ( 1 H NMR at 300 or 500 MHz; 13 C NMR at 101 MHz or 176 MHz). Chemical shifts for 1 H NMR spectra are recorded in parts per million with the residual protic solvent resonance as the internal standard (CDCl 3 : δ = 7.27 ppm). Data are reported as follows: chemical shift [multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, p = pentet, m = multiplet and bs = broad singlet), coupling constant (in Hz), integration and assignment]. 13 C NMR spectra were recorded with complete proton decoupling. Chemical shifts are reported in parts per million with the residual protic solvent resonance as the internal standard (CDCl 3 : δ = 77.0 ppm). Assignments of 1 H and 13 C spectra were based upon the analysis of δ and J values, as well as DEPT, COZY and HSQC experiments where appropriate. All reactions involving air sensitive reagents were performed under nitrogen in oven-dried glassware using syringeseptum cap technique. All organolithiums are commercially available by Sigma Aldrich and were titrated before use. All other chemicals were commercially available and used without further purification. Enantiomeric excess was assessed by HPLC (Chiralcel ODH, ADH). Diastereomeric ratio was assessed by GC-MS or 1 H NMR analysis on the reaction crude.
According to procedure reported in literature (Starmans et al., 1998

1-benzyl-N-(1-hydroxy-2-methylpropan-2yl)azetidine-2-carboxamide 3a
General procedure A: To a perfectly dry flask charged with LaCl 3 (587 mg, 2.40 mmol) was dried by heating under reduced pressure. Then, toluene dry (50 mL) and 2-amino-2methylpropan-1-ol (4.11 g, 46.13 mmol) were added and after cooling to 0 • C, n-hexyllithium (2.0 M in hexane, 23.0 mL, 46.13 mmol) was added dropwise and the reaction was heated to 100 • C for 15 min. The solution was cooled to room temperature and methyl 1-benzylazetidine-2-carboxylate 2a (3.78 g, 18.45 mmol) was added and the mixture was stirred over night. The reaction was quenched with water (1 mL) and filtered over celite. The organic phase was dried over sodium sulfate and the solvent was evaporated under reduced pressure to afford the product as a brown oil (4.50 g, yield 93%).

HYDROLYSIS OF 1,3-OXAZOLIDINES
General Procedure: To a solution of oxazolidines 5a-d, 6a-c and 7b (0.16 mmol, 1 eq) in dichloromethane, silica (150 mg) was added. The reaction was stirred for 3 h at room temperature. The crude was filtered and the solvent was evaporated under reduced pressure to obtain the desire product.
1,3-Oxazolidines 7a and 7c undergo quantitative hydrolysis to corresponding acyl derivatives 10a and c during the chromatography on silica gel.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript/Supplementary Files.

AUTHOR CONTRIBUTIONS
PM, MC, and FF have been involved in the synthesis of all compounds with the help of CC. AA and FF performed X-Ray analysis for compound (S,S)-3b. LP performed DFT and NMR calculations. RL and LD supervised this work and wrote the paper.