Cobalt-Catalyzed Cyclization of 2-Bromobenzamides with Carbodiimides: A New Route for the Synthesis of 3-(Imino)isoindolin-1-ones

A novel synthetic pathway to approach 3-(imino)isoindolin-1-ones by the Co-catalyzed cyclization reaction of 2-bromobenzamides with carbodiimides has been developed. This catalytic reaction can tolerate a variety of substituents and provide corresponding products in moderate yields for most cases. According to the literature, the reaction mechanism is proposed through the formation of a five-membered aza-cobalacycle complex, which carries out the following reaction subsequence, including nucleophilic addition and substitution, to furnish the desired structures.


Introduction
3-(Imino)isoindolin-1-one, an interesting N-heterocycle with two unsymmetrically unsaturated C-X functional groups on the five-membered indole core, has attracted attention from chemists in various fields due to its uncommon structure. However, this unique structure has also enhanced the difficulty of its synthetic pathways; thus, only limited reports have described synthetic methods to approach the derivatives of 3-(imino)isoindolin-1one [1][2][3][4][5][6][7][8][9][10][11][12][13][14]. Among the reported methodologies, transition-metal-catalyzed cyclization reactions represent highly efficient processes that allow synthesis with fewer steps and higher yields. This progress, accompanied by higher efficient synthesis, is significantly beneficial for providing related analogs and able to satisfy the different structural requirements in various fields, which is reported to improve the potential of industrial applications.
For the reported transition-metal catalytic cyclization, the most frequently adopted principle is through the formation of a five-membered metallacycle as the key intermediate. This five-membered intermediate can easily combine with subsequent steps to create diverse N-heterocycles (Scheme 1). A variety of late-transition-metals have been shown to possess the ability to form this intermediate for further transformations [5,7,[15][16][17][18][19][20][21][22][23][24], which have been commonly indicated to construct six-or five-membered N-heterocyclic structures via the 1,2-or 1,1-insertion of molecules containing multiple bonds. For those revealed catalysts, the first-row transition-metal complexes exhibit comparably higher economic efficiency, and our continuing studies on cobalt-catalyzed coupling reactions [25], as well as the experiences in cyclization reactions [26][27][28][29][30][31][32][33][34][35], encouraged us to attempt the synthesis of diverse N-heterocyclic scaffolds by using the five-membered cobalacycle intermediate and switching different cyclization partners (Scheme 2). Under this reaction concept, we were able to obtain several N-heterocyclic compounds, including indenoisoquinolinones, chiral 1-aminoindenes, and isoquinolones. Further study by surveying synthesis of diverse N-heterocyclic scaffolds by using the five-membered cobalacycle in termediate and switching different cyclization partners (Scheme 2). Under this reaction concept, we were able to obtain several N-heterocyclic compounds, including indenoiso quinolinones, chiral 1-aminoindenes, and isoquinolones. Further study by surveying car bodiimide as the coupling partner under a similar reaction protocol allowed us to obtain 3-(imino)isoindolin-1-ones. Scheme 1. Cyclization through the five-membered metallacycle.

Results and Discussion
We carried out the cobalt-catalyzed cyclization reaction of 2-bromobenzamides with alkynes [35], then attempted to replace alkynes with carbodiimides to investigate the re activity of this replacement. Our study started from the optimization of the reaction con ditions by using the model substrates 2-bromo-N-methylbenzamide (1a) and N,N'-dicy clohexylcarbodiimide (2a) with various cobalt catalysts and bases. The results are sum marized in Table 1. We first surveyed the optimized condition of our previous reaction [35]; however, the desired product 3a was obtained in only 47% yield (Entry 1). Although the reaction proceeded smoothly, this result was still unsatisfying and thus made us fur ther investigate various factors of the conditions for this cobalt catalytic cyclization reac tion. We then examined other cobalt catalysts (Entries 2-4) and found that Co(dppe)Cl had the best performance and could provide the desired product 3a in 73% yield (Entry 3). Product 3a was verified by 1 H, 13 C NMR, and high-resolution mass spectrometry synthesis of diverse N-heterocyclic scaffolds by using the five-membered cobalacycle in termediate and switching different cyclization partners (Scheme 2). Under this reaction concept, we were able to obtain several N-heterocyclic compounds, including indenoiso quinolinones, chiral 1-aminoindenes, and isoquinolones. Further study by surveying car bodiimide as the coupling partner under a similar reaction protocol allowed us to obtain 3-(imino)isoindolin-1-ones. Scheme 1. Cyclization through the five-membered metallacycle.

Results and Discussion
We carried out the cobalt-catalyzed cyclization reaction of 2-bromobenzamides with alkynes [35], then attempted to replace alkynes with carbodiimides to investigate the re activity of this replacement. Our study started from the optimization of the reaction con ditions by using the model substrates 2-bromo-N-methylbenzamide (1a) and N,N'-dicy clohexylcarbodiimide (2a) with various cobalt catalysts and bases. The results are sum marized in Table 1. We first surveyed the optimized condition of our previous reaction [35]; however, the desired product 3a was obtained in only 47% yield (Entry 1). Although the reaction proceeded smoothly, this result was still unsatisfying and thus made us fur ther investigate various factors of the conditions for this cobalt catalytic cyclization reac tion. We then examined other cobalt catalysts (Entries 2-4) and found that Co(dppe)Cl had the best performance and could provide the desired product 3a in 73% yield (Entry 3). Product 3a was verified by 1 H, 13 C NMR, and high-resolution mass spectrometry Scheme 2. Cobalt-catalyzed cyclization reactions.

Results and Discussion
We carried out the cobalt-catalyzed cyclization reaction of 2-bromobenzamides with alkynes [35], then attempted to replace alkynes with carbodiimides to investigate the reactivity of this replacement. Our study started from the optimization of the reaction conditions by using the model substrates 2-bromo-N-methylbenzamide (1a) and N,N'dicyclohexylcarbodiimide (2a) with various cobalt catalysts and bases. The results are summarized in Table 1. We first surveyed the optimized condition of our previous reaction [35]; however, the desired product 3a was obtained in only 47% yield (Entry 1). Although the reaction proceeded smoothly, this result was still unsatisfying and thus made us further investigate various factors of the conditions for this cobalt catalytic cyclization reaction. We then examined other cobalt catalysts (Entries 2-4) and found that Co(dppe)Cl 2 had the best performance and could provide the desired product 3a in 73% yield (Entry 3). Product 3a was verified by 1 H, 13 C NMR, and high-resolution mass spectrometry, which was further determined as E-form by single-crystal X-ray analysis (see Supplementary Material for the spectral data and Scheme 3 for the X-ray structure). Additionally, the effect of the base was investigated. When we introduced other bases such as pyrolidine, pyridine, and K 2 CO 3 to replace triethylamine, the yields of the desired product decreased significantly with organic bases (Entries 5 and 6), and the reaction did not proceed with K 2 CO 3 (Entry 7). Next, we tested various solvents to check the reactivity of reactions in different solvents (Entries 8-13). It was found that the yields of product 3a in other solvents were much worse than that in acetonitrile. We often observed the protonation of substrate 1a as the major side product, except the reactions in DMSO. The crude 1 H NMR spectra of the reactions in DMSO were wholly different from those in other solvents. We observed many unidentified side reactions in DMSO, and product 3a could not be clearly isolated. An increase in reaction time was not helpful. When we increased the reaction time from 16 to 24 h, we obtained product 3a in almost the same yield (Entry 14). The blank reactions were investigated as well, and the results indicate that the cobalt catalyst, the ligand, and the reducing agent zinc are key factors and are all necessary (Entries 15-17). The reaction could not proceed without any of them. which was further determined as E-form by single-crystal X-ray analysis (see Supplementary Material for the spectral data and Scheme 3 for the X-ray structure). Additionally, the effect of the base was investigated. When we introduced other bases such as pyrolidine, pyridine, and K2CO3 to replace triethylamine, the yields of the desired product decreased significantly with organic bases (Entries 5 and 6), and the reaction did not proceed with K2CO3 (Entry 7). Next, we tested various solvents to check the reactivity of reactions in different solvents (Entries 8-13). It was found that the yields of product 3a in other solvents were much worse than that in acetonitrile. We often observed the protonation of substrate 1a as the major side product, except the reactions in DMSO. The crude 1 H NMR spectra of the reactions in DMSO were wholly different from those in other solvents. We observed many unidentified side reactions in DMSO, and product 3a could not be clearly isolated. An increase in reaction time was not helpful. When we increased the reaction time from 16 to 24 h, we obtained product 3a in almost the same yield (Entry 14). The blank reactions were investigated as well, and the results indicate that the cobalt catalyst, the ligand, and the reducing agent zinc are key factors and are all necessary (Entries 15-17). The reaction could not proceed without any of them. After obtaining the optimized reaction conditions, we then investigated the reaction scope by testing various substituents to understand the capacity of this cobalt-catalyzed cyclization reaction, as illustrated in Scheme 3. We first investigated the effect of substituents that attached on the amide group (3a-3f). As indicated, primary alkyl groups can provide much higher yields than secondary alkyl and phenyl groups. Thus, the products with N-methyl (3a) and N-methylenefuryl (3b) groups were obtained in 73% yields, but After obtaining the optimized reaction conditions, we then investigated the reaction scope by testing various substituents to understand the capacity of this cobalt-catalyzed cyclization reaction, as illustrated in Scheme 3. We first investigated the effect of substituents that attached on the amide group (3a-3f). As indicated, primary alkyl groups can provide much higher yields than secondary alkyl and phenyl groups. Thus, the products with N-methyl (3a) and N-methylenefuryl (3b) groups were obtained in 73% yields, but the products with N-isopropyl (3c), N-cyclopropyl (3d), N-1-phenylethyl (3e), and N-phenyl (3f) groups could be provided only in 24, 33, 41, and 30% yields, respectively. We next screened the substituents on the benzamide moiety (3g-3m) and found an interesting trend of the reaction. The reaction proceeded well for the electron-donating groups on the backbone phenyl moiety and obtained the corresponding products 3g-3i in moderate yields. However, when an electron-withdrawing group was introduced to the N-methyl substrate, the reactions could not proceed well to form the desired products 3j and 3k. We further modified the N-protected groups from methyl to benzyl to increase the electron density of the amide nitrogen, and the corresponding products 3l and 3m could be obtained in 51 and 43% yields, respectively. These results clearly indicate that the lower electron density on the amide nitrogen led to the lower yields of the desired products.
We further changed the substituent on carbodiimide from the cyclohexyl group to the isopropyl group (3n-3r) to study the performance of this substrate in this cobalt catalytic cyclization reaction. It was found that N,N'-diisopropylcarbodiimide could also proceed with the reaction smoothly when the proper substituents were constructed on substrate 1, and we could observe the same reaction trend through the yields of desired products. Products with the electron-donating groups on the nitrogen or the phenyl moiety (3o and 3q) could provide higher yields than other structures (3n, 3p, and 3r). Moreover, we combined the methoxy group on the phenyl moiety with other N-attached substituents to establish various structures (3s-3v), since an electron-donating group is a guaranteed group to provide the products with comparably higher yields. The products were indeed obtained in moderate yields either with the N-alkyl or the N-aryl groups. Notably, comparing the spectral data of previous reports [5][6][7][8][9][10][11][12][13][14] with ours, as well as the single-crystal structure of 3a, all products were determined as the E-form single isomer as their exact structures.
From the results of the investigation of the substrate scope, we noticed that the reactions strongly rely on the electron density of benzamides. The reactions generally could proceed well only for the electron-rich substrates. For the substrates with an electronwithdrawing group on the backbone phenyl moiety, reactions did not provide satisfying yields or even be inhibited, especially for those with an N-methyl group. This aroused our curiosity and prompted us to study the reaction in detail. We then conducted several control experiments and tried to verify the reaction pathway from the results (Scheme 4). First, we carefully checked the reactions of the electron-deficient substrates 1j and 1k and found that the protonation of substrate 1 dominated the reactions. Compounds 4j and 4k were identified as major products in the reactions, and the recovered yields were very high. No other side reactions were found in these reactions. This implies that the activation of the C-Br bond in these two substrates is fast and clear, and no other competitive reaction can occur in the presence of 1j and 1k. However, when the substrate with an N-methylene-2-pyridinyl group (1s) was utilized in the reaction, the performance of the reaction was totally different. We obtained messy crude NMR spectra. The reaction was complicated, and no compound could be identified as the major product. Several compounds, including the desired product 3w, the corresponding protonation product 4s, and the unidentified side products, could be detected by GC-MS in poor yields. This result was probably caused by the formation of intermediate 1s-A, which was supposed to be formed by the chelation of two nitrogen atoms on substrate 1s. According to the reports, the formation of 1s-A is favorable than that of 1s-A [11][12][13]19], and complex 1s-A will further convert to 1s-B via activation of a C-Br bond. Therefore, the different intermediates resulted in different reaction behavior.  Based on the above results and a previous report [35], we propose the reaction mechanism of this cobalt-catalyzed cyclization reaction as below (Scheme 5). The reaction is likely to be initiated by reduction of the Co(II) catalyst to form an active Co(I) species, which activates the C-Br bond of substrate 1 via oxidative addition or a two-times singleelectron-transfer process to form the resulting Co(III) complex A [20]. An equilibrium between complexes A and A′ will occur by the participation and departure of an HX molecule. Formation of the five-membered aza-cobalacycle complex A′ can provide extra stability for this Co(III) species [11][12][13][14][15], and protonation of complex A will lead to the side product 4. Substrate 2 then participates in the reaction via coordination with complex A, which causes the following nucleophilic addition of the amide group to the central carbon of carbodiimide to obtain another Co(III) complex B. The aryl moiety attached to the Co(III) species is able to perform as a nucleophile to attack the same carbon and generate product 3 with the Co(III) species [15]. Reduction of the resulting Co(III) species leads to the regeneration of the active Co(I) catalyst.
An alternative pathway should be described as proceeding the cyclization by the formation of complex C through the 1,2-insertion of carbodiimide (Scheme 6). However, this reaction pathway will provide product 3 and the undetected compounds 3′ via the nucleophilic substitution of nitrogen to the carbonyl group or the iminyl group, which does not match our results. We did not observe compounds 3′ in the present cobalt catalytic cyclization reaction. Therefore, the formation of complex C is not favorable under the current reaction protocol.

Scheme 4. Control experiments.
Based on the above results and a previous report [35], we propose the reaction mechanism of this cobalt-catalyzed cyclization reaction as below (Scheme 5). The reaction is likely to be initiated by reduction of the Co(II) catalyst to form an active Co(I) species, which activates the C-Br bond of substrate 1 via oxidative addition or a two-times single-electrontransfer process to form the resulting Co(III) complex A [20]. An equilibrium between complexes A and A will occur by the participation and departure of an HX molecule. Formation of the five-membered aza-cobalacycle complex A can provide extra stability for this Co(III) species [11][12][13][14][15], and protonation of complex A will lead to the side product 4. Substrate 2 then participates in the reaction via coordination with complex A, which causes the following nucleophilic addition of the amide group to the central carbon of carbodiimide to obtain another Co(III) complex B. The aryl moiety attached to the Co(III) species is able to perform as a nucleophile to attack the same carbon and generate product 3 with the Co(III) species [15]. Reduction of the resulting Co(III) species leads to the regeneration of the active Co(I) catalyst.
An alternative pathway should be described as proceeding the cyclization by the formation of complex C through the 1,2-insertion of carbodiimide (Scheme 6). However, this reaction pathway will provide product 3 and the undetected compounds 3 via the nucleophilic substitution of nitrogen to the carbonyl group or the iminyl group, which does not match our results. We did not observe compounds 3 in the present cobalt catalytic cyclization reaction. Therefore, the formation of complex C is not favorable under the current reaction protocol.

Conclusions
In conclusion, we developed a cobalt catalysis method to approach 3-(imino)isoindolin-1-one derivatives via the cyclization reactions of 2-bromobenzamides with carbodiimides. This catalytic reaction demonstrated tolerance to diverse substituents and could provide the desired products in moderate yields for most cases. The reaction mechanism is proposed through the formation of a five-membered aza-cobalacycle, which proceeds the nucleophilic addition of amide to carbodiimide with subsequent C-C bond formation to generate the desired products. Further studies of other relative cobalt catalytic reactions, as well as their applications, are currently underway.
Supplementary Materials: The following are available online, experimental procedures, spectral data, X-ray Diffraction Analysis of Compound 3a, 1 H and 13 C NMR spectra for products.