Dicarboxylic Acid Monoesters in β- and δ-Lactam Synthesis

A N-(2-methoxy-2-oxoethyl)-N-(phenylsulfonyl)glycine monomethyl ester of the respective dicarboxylic acid was involved in a reaction with imines promoted by acetic anhydride at an elevated temperature. Instead of the initially expected δ-lactam products of the Castagnoli–Cushman-type reaction, medicinally important 3-amino-2-azetidinones were obtained as the result of cyclization, involving a methylene group adjacent to an acid moiety. In contrast, replacing alcohol residue with hexafluoroisopropyl in the same substrate made another methylene group (adjacent to the ester moiety) more reactive to furnishing the desired δ-lactam in the Castagnoli–Cushman fashion.


Introduction
Lactams of various ring sizes (β [1], γ [2], δ [3], ε [4], denoting 4-7-membered rings, respectively, and larger [5]) represent one of the most important heterocyclic moieties employed in medicinal chemistry and drug design [6]. While the size of the lactam ring has a strong bearing on the chemical and physicochemical properties [7] of the lactamcontaining compounds as well as their specific biological activity profile [8], the broadly defined lactam chemical class can be confidently defined as privileged [9], i.e., capable of delivering compounds endowed with diverse biological activities ( Figure 1). This mandates the development and constant broadening of the current arsenal of synthetic methods to access lactam scaffolds, all with feasible substitution patterns for de novo biological target interrogation and the subsequent medicinal chemistry optimization of the hits emerging from biological screening campaigns.
Recently, we demonstrated that, in the well-known synthesis of tetrahydroisoquinolonic acids from imines and homophthalic anhydride (HPA), the latter can be efficiently replaced with 2-(2-methoxy-2-oxoethyl)benzoic acid (2) (HPA monoester) activated by CDI (1,1 -carbonyldiimidazole). This allowed tetrahydroisoquinolonic esters 3 to be obtained and the Castagnoli-Cushman chemistry reaction to be planted onto a new reagent space that did not require the use of cyclic anhydrides [20]. Encouraged by this finding, we proceeded to investigate if the same approach could be extended to dicarboxylic acid monoesters, such as 4, X = NSO 2 Ph (Scheme 1). Herein, we report our findings obtained during the course of this investigation. Recently, we demonstrated that, in the well-known synthesis of tetrahydroisoquinolonic acids from imines and homophthalic anhydride (HPA), the latter can be efficiently replaced with 2-(2-methoxy-2-oxoethyl)benzoic acid (2) (HPA monoester) activated by CDI (1,1′-carbonyldiimidazole). This allowed tetrahydroisoquinolonic esters 3 to be obtained and the Castagnoli-Cushman chemistry reaction to be planted onto a new reagent space that did not require the use of cyclic anhydrides [20]. Encouraged by this finding, we proceeded to investigate if the same approach could be extended to dicarboxylic acid monoesters, such as 4, X = NSO2Ph (Scheme 1). Herein, we report our findings obtained during the course of this investigation.  [20]; (c) Reaction investigated in this work and initially proposed structure reaction product.

Results and Discussion
Our initial attempt to transfer the reaction conditions optimized for the react HPA monoester 2 with imines (Scheme 1b) was not successful. Therefore, we screen a suitable carboxylic acid activation regimen that would lead to an appreciable conv and product yield (Table 1). Among the three other activators tried (oxalyl chlorid fluoroacetic anhydride and acetic anhydride), only acetic anhydride gave the full co sion of 4 and the best isolated yield for the product (66%, entry 4), whose molecular w and spectral characteristics appeared to correspond to the desired δ-lactam prod and demonstrated the predominant formation of one product diastereomer. Notab reaction turned out to be rather sensitive to thermal activation with a sharp decre the yield either when raising (entry 5) or reducing (entries 6-7) the temperature. Rep chlorobenzene with other solvents (entries 8-9) did not improve the yields, while a a base (entry 10) significantly diminished it.  [20]; (c) Reaction investigated in this work and initially proposed structure of the reaction product.

Results and Discussion
Our initial attempt to transfer the reaction conditions optimized for the reaction of HPA monoester 2 with imines (Scheme 1b) was not successful. Therefore, we screened for a suitable carboxylic acid activation regimen that would lead to an appreciable conversion and product yield (Table 1). Among the three other activators tried (oxalyl chloride, trifluoroacetic anhydride and acetic anhydride), only acetic anhydride gave the full conversion of 4 and the best isolated yield for the product (66%, entry 4), whose molecular weight and spectral characteristics appeared to correspond to the desired δ-lactam product 5a and demonstrated the predominant formation of one product diastereomer. Notably, the reaction turned out to be rather sensitive to thermal activation with a sharp decrease in the yield either when raising (entry 5) or reducing (entries 6-7) the temperature. Replacing chlorobenzene with other solvents (entries 8-9) did not improve the yields, while adding a base (entry 10) significantly diminished it.
However, a single-crystal X-ray analysis of the reaction product demonstrated, to our surprise, that it was not the expected δ-lactam 5a but rather the trans-diastereomer of β-lactam product 6a (Scheme 2). While the activation of α-C-H carboxylic acids toward such a reaction by acetic anhydride (in the presence of triethylamine) has been described in the literature [21,22], the formation of the β-lactam in the absence of any base (as in our case) is hitherto undescribed. Moreover, the fact that the use of the base is detrimental to the product yield (vide supra) makes this transformation rather unique.
Two possible reaction pathways can be suggested for this observed transformation (Scheme 3). They both begin with a carboxylic acid activation by acetic anhydride and the formation of mixed anhydride I. This intermediate can then either directly acylate imine to form intermediate II or be converted into a ketene III via AcOH elimination (the latter can be base-promoted by imine or traces of amine from imine decomposition). Ketene III also reacts with imine to provide the intermediate II. The last step is a Mannich-type cyclization involving the methylene group adjacent to carbonyl, resulting in formation of a C-C bond and beta-lactam cycle. The selective formation of β-lactams over δcan be explained by the increased CH acidity of the methylene group closest to the positively charged iminium fragment of intermediate II compared to the second methylene group adjacent to an ester moiety. sion of 4 and the best isolated yield for the product (66%, entry 4), whose molecular weight and spectral characteristics appeared to correspond to the desired δ-lactam product 5a and demonstrated the predominant formation of one product diastereomer. Notably, the reaction turned out to be rather sensitive to thermal activation with a sharp decrease in the yield either when raising (entry 5) or reducing (entries 6-7) the temperature. Replacing chlorobenzene with other solvents (entries 8-9) did not improve the yields, while adding a base (entry 10) significantly diminished it. However, a single-crystal X-ray analysis of the reaction product demonstrated, to our surprise, that it was not the expected δ-lactam 5a but rather the trans-diastereomer of βlactam product 6a (Scheme 2). While the activation of α-C-H carboxylic acids toward such a reaction by acetic anhydride (in the presence of triethylamine) has been described in the literature [21,22], the formation of the β-lactam in the absence of any base (as in our case) is hitherto undescribed. Moreover, the fact that the use of the base is detrimental to the product yield (vide supra) makes this transformation rather unique. Two possible reaction pathways can be suggested for this observed transformation (Scheme 3). They both begin with a carboxylic acid activation by acetic anhydride and the formation of mixed anhydride I. This intermediate can then either directly acylate imine to form intermediate II or be converted into a ketene III via AcOH elimination (the latter can be base-promoted by imine or traces of amine from imine decomposition). Ketene III also reacts with imine to provide the intermediate II. The last step is a Mannich-type cyclization involving the methylene group adjacent to carbonyl, resulting in formation of a C-C bond and beta-lactam cycle. The selective formation of β-lactams over δ-can be explained by the increased CH acidity of the methylene group closest to the positively charged iminium fragment of intermediate II compared to the second methylene group adjacent to an ester moiety.  However, a single-crystal X-ray analysis of the reaction product demonstrated, to our surprise, that it was not the expected δ-lactam 5a but rather the trans-diastereomer of βlactam product 6a (Scheme 2). While the activation of α-C-H carboxylic acids toward such a reaction by acetic anhydride (in the presence of triethylamine) has been described in the literature [21,22], the formation of the β-lactam in the absence of any base (as in our case) is hitherto undescribed. Moreover, the fact that the use of the base is detrimental to the product yield (vide supra) makes this transformation rather unique. Two possible reaction pathways can be suggested for this observed transformation (Scheme 3). They both begin with a carboxylic acid activation by acetic anhydride and the formation of mixed anhydride I. This intermediate can then either directly acylate imine to form intermediate II or be converted into a ketene III via AcOH elimination (the latter can be base-promoted by imine or traces of amine from imine decomposition). Ketene III also reacts with imine to provide the intermediate II. The last step is a Mannich-type cyclization involving the methylene group adjacent to carbonyl, resulting in formation of a C-C bond and beta-lactam cycle. The selective formation of β-lactams over δ-can be explained by the increased CH acidity of the methylene group closest to the positively charged iminium fragment of intermediate II compared to the second methylene group adjacent to an ester moiety. In addition to the novelty of the discovered protocol for β-lactam synthesis, the medicinal importance of the 3-amino-2-azetidinone scaffold comprised by compound 6a is relatively clear. Indeed, it is the core of exemplary antibiotics, such as 7-10 [23] (Figure 2). This motivated us to explore the scope of the new protocol for β-lactams 6 preparation In addition to the novelty of the discovered protocol for β-lactam synthesis, the medicinal importance of the 3-amino-2-azetidinone scaffold comprised by compound 6a is relatively clear. Indeed, it is the core of exemplary antibiotics, such as 7-10 [23] (Figure 2). This motivated us to explore the scope of the new protocol for β-lactams 6 preparation from monoester 4 and various imines using the optimized conditions (Ac 2 O, PhCl, 130 • C, 16 h).  Following from the results presented in Scheme 4, polysubstituted 3-amino-2-azetidinones 6a-o can be synthesized in a modest-to-high yield (up to 90%) and high transdiastereoselectivity (see also Table S1) in case of aldimines (R 3 = H), as confirmed by a single-crystal X-ray analysis of two reaction products (6a and 6m). The reaction appeared to work equally well for aldimines derived from both aliphatic and aromatic amines. Symmetrical hydrazones could also be productively involved in the reaction (cf., products 6km). In the case of (E)-chalcones, the diastereoselectivity of the reaction deteriorated; however, this was not a detriment to the product yield. The respective diastereomers were separated by HPLC and characterized (compound 6j). The notable products are 6n and 6o, which can be viewed as building blocks for further structural complexity buildup via alkyne-azide click chemistry. Following from the results presented in Scheme 4, polysubstituted 3-amino-2-azetidinones 6a-o can be synthesized in a modest-to-high yield (up to 90%) and high trans-diastereoselectivity (see also Table S1) in case of aldimines (R 3 = H), as confirmed by a single-crystal X-ray analysis of two reaction products (6a and 6m). The reaction appeared to work equally well for aldimines derived from both aliphatic and aromatic amines. Symmetrical hydrazones could also be productively involved in the reaction (cf., products 6k-m). In the case of (E)-chalcones, the diastereoselectivity of the reaction deteriorated; however, this was not a detriment to the product yield. The respective diastereomers were separated by HPLC and characterized (compound 6j). The notable products are 6n and 6o, which can be viewed as building blocks for further structural complexity buildup via alkyne-azide click chemistry.
Encouraged by the results obtained with dicarboxylic acid monoester 4, we proceeded to investigate the workability of the new base-free protocol for other carboxylic acids 11a-m in combination with aldimines, aiming to obtain β-lactams 12 (Scheme 5).
It turned out that the sulfonylamino-substituted acetic acids 11a-d were similarly effective in the β-lactam synthesis, furnishing the respective products 12a-d with good yields and a high diastereoselectivity. When replacing the nitrogen group (X = N) with sulfur, sulfone, oxygen, and carbon linkers had a strong effect on the reaction outcome. The thia-linked carboxylic acids (11e-11g) gave β-lactams in good to fair yields except for, surprisingly, the benzoyl-substituted substrate 11h. Sulfone-, oxygen-and carbon-linked substrates did not deliver the desired products, except for 1,3-dithiane substrate 11m, which furnished the diastereomerically pure product 12m in a modest yield.
Having explored the formation of β-lactams from dicarboxylic monoester 4 and carboxylic acids 11a-m, we continued pondering the possibility of forcing substrates such as 4 to react in the Castagnoli-Cushman fashion and furnish δ-lactam products. One possibility we considered would be to increase the C-H acidity of the 'ester arm' of the substrate, thereby making the closure of the six-membered ring more feasible. This could be achieved by placing electron-withdrawing substituents, such as perfluoroalkyl group, in the ester moiety. In our previous work on the surrogate Castagnoli-Cushman reaction of HPA monomethyl ester 2 and its analogs, we observed a notable increase in the reactivity of the 2,2,2-trifluoroethyl (TFE) ester compared to 2 [20]. Similarly, the TFE ester was found to be more reactive towards nucleophilic addition-elimination [24]. Moreover, TFE esters have been used as versatile acylation reagents in various reactions, including transesterification [25], amidation [26], and the kinetic resolution of aliphatic amines [27]. We reasoned that, aside from TFE analog of monoester 4, the hexafluoroisopropyl (HFIP) congener [28] would be even more predisposed to react with the Castagnoli-Cushman (rather than the previously observed) fashion and yield δ-lactams. Hence, we prepared monoesters 13 (TFE) and 14 (HFIP) (see Supplementary Information) and reacted them with N-butyl-1-(4-methoxyphenyl)methanimine (15) in chlorobenzene in the presence of acetic anhydride (Scheme 6). Expectedly, 14 proved more reactive towards imine 15 compared to its TFE counterpart (13) as it required a lower temperature for the reaction to be completed. Both ester products (16 and labile 17) were hydrolyzed to their respective carboxylic acids (18 and 19), and the spectral characteristics of the latter two compounds were compared to each other to reveal that compound 16 was the product of β-lactam synthesis, while compound 17 was the desired δ-lactam formed via the Castagnoli-Cushman reaction (Scheme 5). Indeed, the signals of the vicinal methine protons in 18 resonated differently (doublets at 4.81 and 4.40 ppm) compared to the same signals in 19 (doublets at 5.06 and 4.85 ppm). Small 3 J coupling constants (2.0-2.2 Hz) were indicative of the trans-configuration of both carboxylic acids [29]. Encouraged by the results obtained with dicarboxylic acid monoester 4, we proceeded to investigate the workability of the new base-free protocol for other carboxylic acids 11a-m in combination with aldimines, aiming to obtain β-lactams 12 (Scheme 5). It turned out that the sulfonylamino-substituted acetic acids 11a-d were similarly effective in the β-lactam synthesis, furnishing the respective products 12a-d with good yields and a high diastereoselectivity. When replacing the nitrogen group (X = N) with sulfur, sulfone, oxygen, and carbon linkers had a strong effect on the reaction outcome. The thia-linked carboxylic acids (11e-11g) gave β-lactams in good to fair yields except for, surprisingly, the benzoyl-substituted substrate 11h. Sulfone-, oxygen-and carbon-linked substrates did not deliver the desired products, except for 1,3-dithiane substrate 11m, which furnished the diastereomerically pure product 12m in a modest yield.
Having explored the formation of β-lactams from dicarboxylic monoester 4 and carboxylic acids 11a-m, we continued pondering the possibility of forcing substrates such as 4 to react in the Castagnoli-Cushman fashion and furnish δ-lactam products. One possibility we considered would be to increase the C-H acidity of the 'ester arm' of the substrate, thereby making the closure of the six-membered ring more feasible. This could be achieved by placing electron-withdrawing substituents, such as perfluoroalkyl group, in the ester moiety. In our previous work on the surrogate Castagnoli-Cushman reaction of HPA monomethyl ester 2 and its analogs, we observed a notable increase in the reactivity of the 2,2,2-trifluoroethyl (TFE) ester compared to 2 [20]. Similarly, the TFE ester was found to be more reactive towards nucleophilic addition-elimination [24]. Moreover, TFE esters have been used as versatile acylation reagents in various reactions, including transesterification [25], amidation [26], and the kinetic resolution of aliphatic amines [27]. We reasoned that, aside from TFE analog of monoester 4, the hexafluoroisopropyl (HFIP) congener [28] would be even more predisposed to react with the Castagnoli-Cushman (rather than the previously observed) fashion and yield δ-lactams. Hence, we prepared  Considering the importance of the notable reactivity switch between monoesters 13 and 14 (β-lactam synthesis vs. Castagnoli-Cushman δ-lactam synthesis), we continued scrutinizing the differences in the structures of products 18 and 19 using NMR spectroscopy after our attempts to obtain crystals suitable for X-ray crystallography failed. The unequivocal difference between these compounds was identified in their correlational HMBC spectra. Specifically, compound 18 displayed two key correlations, 3   Considering the importance of the notable reactivity switch between monoesters 13 and 14 (β-lactam synthesis vs. Castagnoli-Cushman δ-lactam synthesis), we continued scrutinizing the differences in the structures of products 18 and 19 using NMR spectroscopy after our attempts to obtain crystals suitable for X-ray crystallography failed. The unequivocal difference between these compounds was identified in their correlational HMBC spectra. Specifically, compound 18 displayed two key correlations, 3  Considering the importance of the notable reactivity switch between monoesters 13 and 14 (β-lactam synthesis vs. Castagnoli-Cushman δ-lactam synthesis), we continued scrutinizing the differences in the structures of products 18 and 19 using NMR spectroscopy after our attempts to obtain crystals suitable for X-ray crystallography failed. The unequivocal difference between these compounds was identified in their correlational HMBC spectra. Specifically, compound 18 displayed two key correlations, 3   Finally, we reasoned that β-lactam carboxylic acid 18 could be synthesized via the reaction of monomethyl ester 4 followed by the hydrolysis of ester 6p. Similarly, δ-lactam carboxylic acid 19 could be obtained via the Castagnoli-Cushman reaction of dicarboxylic acid 20 mediated by the in situ generation of the respective cyclic anhydride [18] (Scheme 7). To our delight, this synthetic strategy indeed led to the compounds whose spectral Finally, we reasoned that β-lactam carboxylic acid 18 could be synthesized via the reaction of monomethyl ester 4 followed by the hydrolysis of ester 6p. Similarly, δ-lactam carboxylic acid 19 could be obtained via the Castagnoli-Cushman reaction of dicarboxylic acid 20 mediated by the in situ generation of the respective cyclic anhydride [18] (Scheme 7). To our delight, this synthetic strategy indeed led to the compounds whose spectral characteristics fully matched those of compounds 18 and 19 synthesized as described in Scheme 6.

Conclusions
With the aim of involving a dicarboxylic acid monomethyl ester in the recently described Ac2O-promoted Castagnoli-Cushman-type reaction, we identified a novel protocol for β-lactam synthesis instead. A series of 25 novel compounds were prepared in 35-90% yields, mostly as single trans-diastereomers as confirmed by X-ray analysis. The type of substituent in the β-position in the carboxylic acid group of the monoester was found to be crucial for the reaction outcome, while the variation in the imine component was well-tolerated. Additionally, it was discovered that replacing the monomethyl ester with its hexafluoroisopropyl congener not only led to a reduction in the reaction temperature, but also to a marked reactivity switch as the reaction proceeded along the Castagnoli-Cushman-type pathway and furnished the respective δ-lactam. An investigation of the scope of the latter reaction is currently underway in our laboratories and will be reported on in due course.

Conclusions
With the aim of involving a dicarboxylic acid monomethyl ester in the recently described Ac 2 O-promoted Castagnoli-Cushman-type reaction, we identified a novel protocol for β-lactam synthesis instead. A series of 25 novel compounds were prepared in 35-90% yields, mostly as single trans-diastereomers as confirmed by X-ray analysis. The type of substituent in the β-position in the carboxylic acid group of the monoester was found to be crucial for the reaction outcome, while the variation in the imine component was well-tolerated. Additionally, it was discovered that replacing the monomethyl ester with its hexafluoroisopropyl congener not only led to a reduction in the reaction temperature, but also to a marked reactivity switch as the reaction proceeded along the Castagnoli-Cushman-type pathway and furnished the respective δ-lactam. An investigation of the scope of the latter reaction is currently underway in our laboratories and will be reported on in due course.

General Procedure for Preparation of Beta Lactams 6a-p, 12a-m, 16 and Their Analytical Data
In a screw-cap vial equipped with a magnetic stir bar, imine (1.1 eq) and the corresponding substituted monocarboxylic acid (0.05-1.3 mmol) were mixed in chlorobenzene (0.058 M, 1-6 mL). Then, acetic anhydride (1.1 eq) was added. The resulting mixture was placed in a pre-heated to 130 • C oil bath or metal heating block. After 16h, the mixture was cooled to room temperature, and the solvent was evaporated. The residue was purified by column chromatography in silica gel with a linear gradient (5-75%) of acetone in hexane (total volume of eluent, 400 mL) to provide pure compounds, 6a-p and 12a-m.  13