Cooperative Al(Salen)-Pyridinium Catalysts for the Asymmetric Synthesis of trans-Configured β-Lactones by [2+2]-Cyclocondensation of Acylbromides and Aldehydes: Investigation of Pyridinium Substituent Effects

The trans-selective catalytic asymmetric formation of β-lactones constitutes an attractive surrogate for anti-aldol additions. Recently, we have reported the first catalyst which is capable of forming trans-β-lactones with high enantioselectivity from aliphatic (and aromatic) aldehyde substrates by cyclocondensation with acyl bromides. In that previous study the concepts of Lewis acid and organic aprotic ion pair catalysis were combined in a salen-type catalyst molecule. Since a pyridinium residue on the salen periphery is essential for high trans- and enantioselectivity, we were interested in the question of whether substituents on the pyridinium rings could be used to further improve the catalyst efficiency, as they might have a significant impact on the effective charges within the heterocycles. In the present study we have thus compared a small library of aluminum salen/bispyridinium catalysts mainly differing in the substituents on the pyridinium residues. As one result of these studies a new catalyst was identified which offers slightly superior stereoselectivity as compared to the previously reported best catalyst. NBO calculations have revealed that the higher stereoselectivity can arguably not be explained by the variation of the effective charge.


Introduction
β-Lactones (systematic name 2-oxetanones), are attracting the interest of scientists for mainly two reasons: (1) a number of natural and synthetic β-lactones are known to act as specific enzyme inhibitors [1][2][3][4]. Tetrahydrolipstatin (Xenical ® , orlistat), for instance, is used for the treatment of obesity and has recently received renewed attention due to the finding that it is capable of specifically inhibiting fatty acid synthase (FAS-TE), an approved drug target for cancer treatment [5][6][7]; (2) as a result of their inherent ring strain [8] β-lactones are useful synthetic building blocks. Ring opening with hard nucleophiles offers the possibility to a divergent access to aldol products [9][10][11][12][13], whereas treatment with soft nucleophiles can be utilized to synthesize β-functionalized carboxylic acid derivatives [9,10,12]. In both cases the stereoinformation of the β-lactone ring can be completely transferred into the ring opening product and the acyl-oxygen bond cleavage with hard nucleophiles typically proceeds with retention of configuration. Cis-and trans-configured β-lactones thus behave as masked and activated syn-and anti-aldol equivalents, respectively.
Recently, we have reported a conceptually new approach which is applicable to both aliphatic and aromatic aldehydes [53,54]. Aluminum salen complexes carrying aprotic organic ion pairs in the periphery of the salen core were investigated. The most useful described catalyst 4aA in terms of enantioselectivity and reactivity was equipped with two pyridinium bromide functionalities connected via benzylic methylene linkers to the ortho-position of the phenolate O atoms (3/3'-position, Scheme 1).

Scheme 1.
Application of bispyridinium Al-salen complex 4aA for the trans-selective catalytic asymmetric synthesis of 3,4-disubstituted 2-oxetanones 3 [53,54]. R  By the choice of a Lewis acid with only one available coordination site a cyclic Zimmerman-Traxler-type transition state leading to the cis-configured product would be avoided. Our catalyst combines the cooperative action of a Lewis acid and of an aprotic onium halide ion pair Q + X − (in the case of 4aA pyridinium bromide), the latter being arguably used to generate the acyl halide enolate 6 in the catalyst sphere (Scheme 2). Q + might stabilize the otherwise unstable enolate [55][56][57][58] by ion pair formation to increase the lifetime of the enolate species and to direct the enolate to the aldehydeLewis acid complex 7 via an open transition state 8 adopting a staggered conformation around the generated C-C bond [54]. In the reactive conformation the disubstituted enolate C-1 atom should be oriented gauche to the aldehyde function's H atom and the C-2-enolate-H atom would be expected to direct toward the sterically demanding Lewis acid-ligand complex, so as to minimize repulsive interactions. The initial aldol adduct could then cyclize to form the heterocyclic product 3.
The general catalyst principle, utilizing the cooperative action of a Lewis acid and an aprotic organic ion pair, is illustrated in Figure 1. Cooperative action of a Lewis acid and an aprotic organic ion pair within a single catalyst entity: the Lewis-acid serves to activate an electrophile, whereas the in situ generated anionic nucleophile forms an ion pair with an aprotic organic cation (Q + ), which stabilizes and directs the nucleophile towards the electrophile [53,54].
Salen ligands were chosen due to their modular nature and their ready and rapid accessibility [59][60][61]. As a result of the proposed stabilizing electrostatic interaction of Q + and the acyl halide enolate 6, the charge density of Q + was expected to play a decisive role on the reaction outcome, in particular in terms of trans/cis-diastereoselectivity and enantioselectivity, because neutral substituents at the 3/3'-position lead predominantly to the undesired almost racemic cis-diastereomer [53,54].
In our continuing efforts to utilize ketenes [62][63][64][65] and related reactive intermediates [66][67][68][69][70][71] for a rapid and practical access to chiral building blocks we have therefore investigated a series of Al-salen catalysts 4 ( Figure 2) carrying different substituents Z on the pyridinium moiety (acting as Q + ) to study electronic effects on the catalyst performance. The results are compared to NBO calculations which have revealed the charge distribution within the pyridinium rings.

Ligand Preparation
In our previous studies we found that a major limitation of catalyst 4aA results from its relatively low solubility at low reaction temperatures. For that reason we were also interested in derivatives in which the phenolic t-Bu-substituents have been formally replaced by n-pentyl groups in order to increase the rotational freedom of the alkyl substituents R 1 and hence the catalyst solubility. To investigate the steric influence of the R 1 substituent on the stereoselectivity, R 1 = Me was also studied.
Based on our previously established protocol [53,54] the ligands 14 could be prepared via a general 4-step procedure (Scheme 3). Phenols 9 were formylated with paraformaldehyde in the presence of NEt 3 and MgCl 2 [72,73]. A subsequent bromomethylation of 10 with paraformaldehyde, HBr and catalytic amounts of sulfuric acid gave benzyl bromides 11 in high yields [74]. The latter were then treated with different commercially available pyridine derivatives in a nucleophilic substitution reaction in acetonitrile (Table 1). The pyridinium salts 12 could be isolated and purified by precipitation and washing with diethyl ether. Nucleophilic substitution proceeded smoothly in most cases. Only the 4-Cl derivative 12bH was prone to releasing the starting pyridine in the back reaction (entry 11). This is ascribed to the electron-withdrawing character of the Cl-substituent, which reduces the nucleophilicity of the pyridine N and leads to a more potent leaving group. The chloro-and iodo-substituents at the 4-position of the pyridine were also found to be sensitive to a partial halogen exchange with bromide (entries 10 and 11).

Scheme 3.
Synthesis of a salen ligand library 14 with different residues R 1 and R 2 .  Treatment of two equivalents of the corresponding aldehyde 12 with enantiopure (R,R)-1,2diaminocyclohexane (13) in EtOH at room temperature for 15 h in the presence of 4 Å molecular sieves provided the salen ligands 14 (Scheme 3 and Table 1). The diimine formation from aldehydes 12 usually furnished 14 in high yields, with the exception of the 4-chloro-and the 4-cyanopyridine derivatives 14bH and 14bI (entries 11 and 12). In both cases none of the desired ligands was obtained, again arguably due to the pronounced leaving group properties of the more electron poor pyridines.

Catalysis
To examine the impact of the different pyridinium residues R 2 , all accessible ligands were investigated in the cyclocondensation of propionylbromide (1A) and dihydrocinnamaldehyde (2a, Table 2). Due to the air and moisture sensitivity of the investigated aluminum complexes 4, the catalysts were generated in situ from 10 mol% of the corresponding ligand 14 and 10 mol% AlMe 3 in CH 2 Cl 2 at room temperature. We have previously already reported that the yields of β-lactone formation are generally higher with isolated catalysts, but isolation has only a small impact on enantioand diastereoselectivity [53,54]. The catalyst solution was cooled to −70 °C and treated with both reagents and finally with iPr 2 NEt. After stirring for 24 h at −70 °C, the reaction was terminated by addition of hydrochloric acid.  In most cases, conversion of the aldehyde was good and yields of β-lactone 3Aa were moderate to good. Catalysts generated from 14bA and 14bB with R 1 = n-Pent (entries 2 and 5) resulted in higher conversions and yields than their direct counterparts derived from 14aA and 14aB, respectively, carrying a t-Bu residue R 1 (entries 1 and 4). This is attributed to a higher solubility in CH 2 Cl 2 at −70 °C with a more flexible alkyl chain [75]. 14cA with R 1 = Me resulted in poor solubility, explaining the lower yield (entry 3) compared to the results with 14aA and 14bA, respectively (entries 1 and 2). For that reason additional Me derivatives were not studied.
Interestingly, the nature of R 1 was found to have only a minor effect on both enantio-and diastereoselectivity (compare entries 1, 3, 4 and 5), whereas most salen catalyzed reactions require enhanced steric demand at that position for an efficient transfer of chirality [76]. In the present case, the stereoselectivity is much more dependent on the pyridine substituents. The best combination of enantioselectivity (ee = 90%) and diastereoselectivity (trans/cis = 97:3) was attained with ligand 14bB carrying a 4-Me substituent as a weak -donor on the pyridine ring (entry 5). This ligand also allowed for a useful product yield.
Increasing the steric demand of the pyridine by a 4-t-Bu substituent had only a minor impact on the reactivity, but both the dr (92:8) and the ee (80%) were negatively affected (entry 6). Reduced diastereoselectivity data were also noticed with a 3,5-dimethyl substitution pattern on the pyridine (entry 7) or with a quinoline residue (entry 8).
The DMAP derivative 14bF (entry 9) carrying the potent -donor group NMe 2 at the pyridine 4-position and the derivative 14bG (entry 10) carrying an iodo atom as a weak -acceptor at the 4-position resulted in similar reactivity and stereoselectivity. For 14bF, a lower diastereoselectivity was expected as a consequence of the +M-effect which should result in a wider charge distribution thereby weakening the postulated contact ion pair with the acyl halide enolate 6. Consequently, -acceptors like in 14bG should result in a more efficient ion pair formation and improved stereoselectivity, but steric effects as well as the poor solubility of 14bG even at room temperature might have overwritten this effect. Unfortunately, it appears to be difficult to study the effect of -acceptors in ligands 14, as they result in a lower stability of the catalysts.
In contrast, the 4-Me group does not only lead to an improved catalyst stability (the ligand 14bB can be stored for at least two months at room temperature with no decomposition detected) and solubility, but might also have a positive effect on a uniform reactive conformation of the generated acyl halide enolate, whereas larger residues might hamper an efficient ion pair formation of the enolate and the pyridinium residue.
The most stereoselective catalyst 4bB was applied to different substrates (Table 3). For the investigated aldehydes and acyl bromides 4bB gave always slightly higher enantioselectivities compared to our previously reported system 4aA. Moreover, the trans-selectivity was equal (entries 1 and 2) or better (entries 3 and 5) than with catalyst 4aA. Entry 2 shows that improved yields can be obtained with the isolated catalyst 4bB. However, the more convenient protocol with in-situ catalyst formation allows for similar diastereoand enantioselectivity (compare entries 1 and 2). The highest enantioselectivity was attained with the aromatic benzaldehyde (entry 4, ee = 96%), but the yield is significantly lower than for aliphatic aldehydes. This is mainly a consequence of the marked sensitivity towards elimination of 4-aryl substituted 2-oxetanones explaining a partial decomposition during the workup [77].

Mechanistic and Theoretical Investigations
Reactions catalyzed by salen complexes are known to proceed in several cases via bimetallic reaction pathways [78]. As part of our programme on bimetallic cooperative catalysis [79][80][81][82], we were therefore interested if two salen units might also cooperate to form the β-lactone products 3.
In that case the presence of a non-linear effect (NLE) would be expected [78,83]. The absence of a NLE in the present case ( Figure 3) indicates that a major product formation pathway involving two salen molecules is most likely no realistic scenario. Our mechanistic considerations thus focus on reaction pathways involving a single catalyst molecule. As mentioned above, we have previously shown that a cationic residue on the salen periphery is essential for high trans-selectivity, but also for high enantioselectivity and reactivity [53,54]. In the case of standard Al-salen complexes carrying a H atom or an isobutyl residue at the 3/3'-position, reactivity was poor and almost racemic product was formed favoring the cis-isomer. With a t-Bu residue no product was formed at all. In the initial study a pyridinium residue was found to be superior compared to ammonium residues [53,54]. We have tentatively explained this preference by a more efficient contact ion pair formation of the planar enolate moiety with the planar pyridinium system as compared to tetrahedral ammonium moieties.
As the positive charge of the substituent at the 3/3'-position is essential for a successful reaction outcome, we were interested in the effective charges for the derivatives described above. For that reason we accomplished a series of NBO analyses using the MOLPRO package of ab initio programmes [84]. For the electronic structure calculations we chose density functional theory in combination with a double-ζ basis set, i.e., B3LYP/cc-pVDZ. Our original idea was that an acceptor substituent in the 4-position of the pyridinium ring might amplify the effect of the positive charge and could further stabilize the contact ion pair. In agreement with the literature [85], our calculations revealed a negative partial charge for the nitrogen atoms, for both tetramethylammonium and pyridinium cation systems (see Table 4). The positive partial charges in the tetramethylammonium reference are distributed over the nine hydrogen atoms. In contrast the pyridinium cation has positive partial charges on the α-carbon atoms, which might stabilize the enolate in the proposed contact ion pair more efficiently than in the case of quaternary ammonium cations, since the positive charge is wider distributed in the latter case. Pyridinium derivatives with substituents in the 4-position give nearly identical partial charges for the nitrogen and the α-carbon atoms as the parent system with a H atom in 4-position. Only the partial charge of the carbon in the 4-position shows a noticeable change on substitution.
For the cooperative contact ion pair/Lewis acid activation two reaction pathways appear to be feasible (Scheme 4), both leading to the observed absolute and relative β-lactone configurations. The two reaction mechanisms differ in the reactive conformations of the coordinated aldehyde and in the pyridinium unit involved. In path A presented in Scheme 4, the aldehyde group's H atom is expected to point toward the C=N imine bond in 15 connected to phenolate ring A. Pyridinium ring A forms the reactive contact ion pair with the acylbromoenolate. The depicted aldehyde conformation appears to be required to get the aldehyde in close distance to the reactive enolate in a transition state adopting the above proposed staggerd conformation (see Scheme 2) to form the trans-β-lactone with a (3S,4S)-configuration.

Scheme 4. Two plausible reaction pathways via cooperative contact ion pair directed
Lewis acid activation leading to the observed absolute and relative configurationof β-lactones 3.  In path B the same aldehyde H atom would point away from the chiral salen backbone. To form the product with the observed absolute configuration again the Re-face of the aldehyde has to react and this would require the action of the other pyridinium moiety B.
We favor this scenario because the repulsive interactions of the aldehyde and the salen core would be minimized. An aldehyde conformation like in path B is often hampered in salen-complex catalyzed reactions by the presence of tBu groups at the 3/3'-position [76,86]. However, the aldehyde conformation in path A appears more unfavorable in the present case given the folding of chiral salen metal complexes derived from trans-1,2-diaminocyclohexane which describes the ligand deviation from a planar arrangement [87][88][89]. In the present case phenolate ring B is expected, based on literature precedent for related Al salen complexes [90], to fold downward toward the Me-Al bond in a stepped conformation. That means, the B-half offers more available space for substrate accomodation.

General
1 H-NMR and 13 C-NMR spectra were measured on a Bruker Avance spectrometer (300 or 500 MHz, Rheinstetten, Germany) in CDCl 3 or DMSO as solvent using TMS as internal standard and chemical shifts are expressed as δ in ppm. Molecular masses were determined with the electron spray ionization (ESI) method on a MicroTOFQ (Bruker, Bremen, Germany) spectrometer. IR spectra were recorded on a Bruker Vector 22 FT-IR spectrometer (Bremen, Germany) with an ATR module (Golden Gate). Melting points are uncorrected and were measured on a Büchi Melting Point B-535 analysis device. The enantiomeric excesses were determined by HPLC on an Elite LaChrom system equipped with Hitachi modules. A chiral stationary phase Daicel column of the Chiracel OD-H type was used.

General Procedure for the Synthesis of 10 (GP1)
This procedure was according to a published protocol [72,73]. To a solution of the phenol derivative 9 (1 eq., 15.0 mmol) in acetonitrile, paraformaldehyde (6.7 eq.), magnesium chloride (1.5 eq.) and triethylamine (3.8 eq.) were added and the mixture was heated under reflux for 24 h. After cooling to ambient temperature, 1 M hydrochloric acid was added till the yellow residue dissolved, followed by extraction with diethyl ether (3 × 20 mL). The organic layers were dried over MgSO 4 , filtered and the solvent was removed in vacuo. Purification by column chromatography (petroleum ether/ethyl acetate 10:1) gave the aldehyde 10.

General Procedure for the Synthesis of 11 (GP2)
This procedure was according to a published protocol [74]. To the corresponding aldehyde 10 (1 eq., 10.5 mmol) were added aq. hydrobromic acid (48%, 7.5 eq., 78.9 mmol, 8.6 mL), paraformaldehyde (1.5 eq., 15.8 mmol, 0.47 g) and a catalytic amount of sulfuric acid (3 drops). Depending on the alkyl chain R 1 , the mixture was stirred at 70 °C for 1 to 5 days. After cooling to ambient temperature, water (10 mL) was added followed by extraction with methylene chloride (3 × 10 mL). The collected organic layers were dried over Na 2 SO 4 , filtered and the solvent was removed in vacuo.

General Procedure for the Synthesis of 12 (GP3)
This procedure was according to a published protocol [92]. To a solution of 11 (1 eq., 3.0 mmol) in acetonitrile (8 mL) was added the corresponding pyridine derivative (1.1 eq., 3.3 mmol). The mixture was stirred for 15 h at ambient temperature. For the workup the solvent was removed in vacuo till a volume of ca. 5 mL was reached and the product was precipitated with diethyl ether (10 mL). After drying in vacuo the salt 12 was obtained.

General Procedure for the Synthesis of 14 (GP4)
This procedure is according to a published protocol [54]. To a solution of (1R,2R)-(−)-1,2diaminocyclohexane (1 eq., 0.15 mmol, 17.1 mg) in ethanol (0.7 mL) at ambient temperature molecular sieves (4 Å) and the corresponding salt 12 (2 eq., 0.30 mmol) were added. The mixture was stirred for 15 h at ambient temperature. For workup the mixture was filtered and washed with ethanol. The solvent was removed in vacuo till a volume of 5 mL was reached and the ligand was precipitated with diethyl ether (30 mL). The solid product was then dried in vacuo.

General Procedure for the Catalytic Asymmetric Synthesis of 3 (GP5)
To a solution of the salen ligand 14 (0.1 eq., 75 µmol) in CH 2 Cl 2 (3 mL) a solution of Me 3 Al in toluene (2 M, 0.1 eq., 75 µmol, 38 µL) was added and the mixture was stirred for 3 h at ambient temperature. Afterwards aldehyde 1 (1 eq., 0.75 mmol), acylbromide 2 (6 eq., 4.5 mmol) and diisopropylethylamine (2.5 eq., 1.88 mmol) were added at −70 °C and the reaction mixture was stirred for 24 h at this temperature. The reaction was quenched by pouring into aqueous 1 M HCl (30 mL) and the product was extracted with CH 2 Cl 2 (2 × 20 mL). The combined organic layers were dried over MgSO 4 and filtered through a pad of silica gel. After removing the solvent in vacuo the desired transβ-lactone 3 was obtained. (3S,4R)-trans-3-Methyl-4-phenyloxetan-2-one (3Ac, 0,18 mmol, yield 37%, ee = 96%, dr = 98:2) was prepared from propionyl bromide (1A) and benzaldehyde (2c) according to GP5, but using 0.49 mmol of aldehyde in 2 mL of CH 2 Cl 2 and basic workup conditions. The reaction mixture was quenched with diisopropylethylamine (2 mL) and filtered through a short plug of Et 3 N-deactivated silica gel. CH 2 Cl 2 was subsequently removed in vacuo. The crude mixture was purified by flash chromatography (Et 3 N-deactivated silica gel, ethyl acetate/petroleum ether 1:10). The dr value was determined by 1 H-NMR and the ee value by HPLC (Chiralcel OD-H, 99:1 n-hexane/iPrOH, 0.5 mL/min, 210 nm). To a solution of ligand 14bB (0.13 g, 0.16 mmol, 1.0 equiv.) in CH 2 Cl 2 (3.0 mL) a solution of Me 3 Al in toluene (2 M, 0.10 mL, 0.16 mmol, 1.0 equiv.) was added. The mixture was stirred for 3 h at ambient temperature. The reaction mixture was poured into 20 mL of pentane to precipitate complex 4bB. Subsequently the mixture was centrifuged and the supernatant removed. Washing the catalyst with additional pentane (10 mL) and drying in vacuo afforded the active catalyst as orange powder in quantitative yield. C 45

Conclusions
In summary, we have reported a catalyst which offers the highest enantio-and trans-selectivity known so far for the catalytic asymmetric synthesis of β-lactones by [2+2] cyclocondensation of acyl halides and aldehydes. Catalysts for the asymmetric formation of trans-β-lactones are of major interest, since trans-β-lactones offer a divergent and atom-economic access to the important class of anti-aldol products. In our catalyst, an Al-center (offering a single coordination site) cooperates with a picolinium bromide moiety based on our recently published strategy to combine the concepts of Lewis acid and organic aprotic ion pair catalysis in a single catalyst system. Since cationic residues like pyridinium units have been found to be essential for both high trans-and enantioselectivity (suggesting that the positive charge enables an ion pair catalysis pathway), we have investigated the question, if substituents on the pyridinium rings can be utilized to further improve the catalyst efficiency, as they might display a significant impact on the effective charges. In the present study we have thus compared a small library of aluminum-salen/bispyridinium catalysts, mainly differing in the substituents on the pyridinium rings. NBO calculations have revealed though that the different catalyst efficiencies can arguably not be explained by the variation of the effective charges, since there are only very small differences for -donor or -acceptor substituted pyridinium systems. However, we have noticed that the substituents have a major impact on the catalyst stability and presumably they have also an impact on the reactive conformation of the proposed acyl halide enolate intermediates.