Investigation of reaction conditions. We commenced our studies by performing the reaction of N-benzyl quinolinium bromide A1, N-ethyl aniline B1, paraformaldehyde, and base in MeOH at 65 oC for 18 h by employing [RuCl2 (p-cymene)]2 as the catalyst. Among various bases tested, Mg(OMe)2 exhibited the best chemo-selectivity since there is no formation of by-product N-benzyl tetrahydroquinoline A1’’ (Table 1, entries 1-4). The absence of catalyst or base failed to yield product C1 (entries 5-6), showing that both of them are indispensable for the product formation. Then, we screened several other metal catalysts applied frequently in hydrogen transfer reactions (see Table S1 in the Supplementary Information (SI)). The results showed that Ir(I) or Ir(III) catalysts were also applicable, but the base metal catalysts (Co, Fe, Mn, and Ni) were totally ineffective for the transformation (entries 7-8 and Table S1). Here, we chose the cost-effective [Ru(p-cymene)Cl2]2 as the preferred catalyst to further evaluate the solvents and temperatures, it showed that methanol and 55 oC were more preferable (entries 9-10). Decrease of the base or (CH2O)n amount diminished the product yields (entries 11-12). Thus, the optimal yield of product C1 was obtained when the reaction in methanol was performed at 55 oC for 18 h by using the combination of [Ru(p-cymene)Cl2]2 and Mg(OMe)2 (entry 10). Interestingly, the use of Mg(OMe)2 base always resulted in excellent selectivity in affording product C1 (entries 3, and 7-12).
Substrate scope. With the availability of the optimal reaction conditions (Table 1, entry 10), we then assessed the substrate scope of the newly developed synthetic protocol. As shown in Fig. 2, various quinolinium salts A (A1−A21, see Scheme S1 for their structures) in combination with N-ethylaniline B1 and paraformaldehyde were evaluated. Gratifyingly, all the reactions underwent smooth reductive annulation and furnished the desired fused N-heterocycles in reasonable to excellent isolated yields with excellent syn-diastereoselectivity (C1−C21, d.r. > 20 : 1). The structure of compound C1 was confirmed by X-ray crystallography diffraction and NOESY spectrum (Fig. S1-Fig.S3). As expected, the application of 1,5-dibromopentane generated the alkyl-linked product C21 in good yield. Noteworthy, a variety of functionalities (e.g., −Me, −OMe, −SPh, amido, −F, −Cl, −Br, ester, −CF3, −NO2, alkenyl, and alkyl) on the quinolinium salts were well tolerated, and their electronic properties affected the product formation to some extent. Interestingly, no reduction of the nitro and alkenyl groups was observed (C17 and C18), and the halo-substrates also did not undergo hydrodehalogenation, indicating that the reaction proceeds in a chemoselective manner. In general, quinolines bearing an electron-donating group (C2−C7, and C13−C14) afforded relatively higher product yields than those having an electron-withdrawing group (C8−C10, and C12), presumably because the electron-rich quinolinium salts can result in more reactive enamine intermediates that are beneficial to the electrophilic coupling process (Fig. 1a). The retention of these functionalities offers the potential for post-functionalization of the obtained products.
Next, we turned our attention to the synthesis of structurally diversified products by variation of both azaarenes A' and anilines B. First, a series of N-alkyl anilines (B2−B18, see Scheme S2 for their structures) in combination with quinolinium salt A1 were tested. As illustrated in Fig. 3, all the reactions efficiently afforded the desired product in moderate to excellent isolated yields with exclusive syn-selectivity (C22−C36, d.r. > 20 : 1). The electronic properties of the substituents on the benzene ring of the anilines significantly affected the product formation. Especially, anilines containing electron-donating groups (C22−C23, C27 and C35) gave much higher yields than those with electron-withdrawing groups (C24−C25). This observation is attributed to electron-rich anilines favoring the electrophilic coupling process during the formation of the products. In addition to N-alkyl anilines, diarylamine B16 also served as an effective coupling partner, affording the N-aryl product C33 in moderate yield. As expected, primary anilines were not applicable for the transformation, as they easily reacted with formaldehyde to form aminals. Interestingly, tetrahydroquinolines (B8 and B9) and 2,3,4,5-tetrahydro-1H-benzo[b]azepine (B10), two specific aniline derivatives, also underwent efficient multicomponent annulation to afford the polycyclic products (C34−C36, C38 and C41). In addition to quinolines, other azaarenes, such as 1,8-naphthyridines (A22−A25), thieno[3,2-b]pyridine A26, 1,7-phenanthroline A27, 1,10-phenanthroline A28, 5-substituted isoquinolines, and more challenging pyridine were also amenable to the transformation, delivering the desired products in an efficient manner (C37−C50, d.r. > 20 :1), these examples demonstrate the capability of the developed chemistry in the functionalization of pyridine-containing azaarenes including the N-bidentate ligands (C37−C42, C47).
Noteworthy, 5-substituted isoquinolines afforded the desired annulation products (Fig. 3, C48 and C49), whereas 5-nonsubstituted isoquinolines generated structurally novel products D by installing an additional methyl group at the β-site of the N-heteroaryl reactants, and all the products are produced with exclusive syn-diastereoselectivity (d.r. > 20 :1, Fig. S4). As shown in Fig. 4, N-benzyl isoquinolinium salts were firstly employed to couple with paraformaldehyde and N-ethyl aniline B1. All the reactions gave rise to the desired annulation products in moderate to excellent yields upon isolation (D1−D13). Then, the transformation of secondary anilines including the N-alkyl and N-aryl ones was evaluated. Gratifyingly, all these anilines smoothly coupled with N-benzyl quinolium salt A1 and paraformaldehyde, delivering the annulation products in reasonable to high yields (D14−D26). Similar to the results described in Fig. 2 and 3, various functionalities on both isoquinolium salts and anilines are well tolerated (−Bn, −Et, −Me, −F, −Cl, −Br, boronic ester, −SO2Me, −n-hexyl, −OMe, −CF3, −CO2Me, alkenyl, cyclohexyl, and i-propyl). The substituents on the aryl ring of the isoquinoline salts have little influence on the product formation, whereas the substituents of the anilines significantly affected the product yields. Generally, aniline bearing an electron-donating group afforded higher yields (e.g., D14−D16 and D20−D23) than those of anilines with an electron-withdrawing group (e.g., D17−D19 and D24), suggesting that the reaction involves an electrophilic coupling process. Benzocyclic amines (1,2,3,4-tetrahydroquinoxaline, 1,2,3,4-tetrahydroquinoline, and 2,3,4,5-tetrahydro-1H-benzo[b]azepine) and N1-isopropyl-N4-phenylbenzene-1,4-diamine also served as effective coupling partners, affording the polycyclic products in moderate to high yields (D27−D30). These examples show the practicality of the developed chemistry in the construction of structurally complex polycyclic N-heterocycles.
Synthetic applications. Further, we explored the synthetic applications of the developed chemistry. As shown in Fig. 5a, 6-ester quinolinium salts, arising from initial esterification of 6-carboxylic quinoline and subsequent pretreatment with benzyl bromide, efficiently reacted with aniline B1 and paraformaldehyde to afford products C51 and C52, which are the analogues of analgesic48 and the agents used for antioxidation and antiproliferation49, respectively. Through successive amidation and formation of N-benzyl heteroarenium salt, 6-amino quinoline was efficiently transformed in combination with aniline B1 into camphanic amide C53 (Fig. 5b), an agent capable of stereoisomeric separation.50 Further, the gram-scale synthesis of product D1 was successfully achieved by scaling up the reactants to 10 mmol, which still gave a desirable yield (Fig. 5c). Interestingly, representative compounds C29 and D1 underwent efficient debenzylation to afford N-unmasked products C54 and D31 in the presence of a Pd/HCOONH4 system in methanol (Fig. 5d), which demonstrates the practicality of the developed chemistry in further preparation of fused heterocycles containing a useful -NH motif.
Mechanistic investigations. To gain mechanistic insights into the reaction, we conducted several control experiments (Fig. 6). First, the model reaction was interrupted after 6 hours to analyze the product system. Except for the formation of product C1 in 23% yield, a dihydroquinoline int-1 was isolated in 5% yield (Fig. 6a). Subjection of compound int-1 (Fig. 6a and Fig. S6) with aniline B1 under the standard conditions resulted in product C1 in high isolated yield (Fig. 6b), showing that int-1 is a key reaction intermediate. However, removal of Ru-catalyst from the standard conditions failed to produce C1 and the α-arylated product C1' (Fig. 6c), revealing that the reaction initiates with Ru-catalyzed hydrogen transfer, instead of nucleophilic arylation of substrate A1 with aniline B1. Further, the model reaction using deuterated methanol solvent yielded product C1 without any D-incorporation (Fig. 6d). In sharp contrast, the same reaction by replacing paraformaldehyde with the fully deuterated one gave product C1-dn with 35% and 27% D-ratios at the α and γ-sites and more than 99% D-ratio at the newly formed aminomethyl group (Fig. 6e and Fig. S8). These two crucial experiments show that the formaldehyde serves as both the source of the reductant and C1-building block for the formation of the newly formed β-methylene group, and there is a tautomerism between int-1 and enamine int-2 (Fig. 1a). In parallel, we conducted the control experiments in terms of the generation of product D1 (Scheme S3). The results also support that dihydroquinoline int-6 (Scheme S3b) and β-methyl dihydroquinoline int-9 (Fig S7) are the reaction intermediates, and formaldehyde serves as the reductant source and C1-building block in the construction of the product (Scheme S3d, S3e and Fig. S9).
Based on the above findings, the plausible pathways toward the formation products C1 and D1 are depicted in Fig. 7. Initially, the metal hydride species [RuIIHX] is generated via Mg(OMe)2 addition to formaldehyde (E1) followed by transmetallation (E2) with [RuIIX2] and β-hydride elimination and release of formate ester (detected by GC-MS analysis, Fig. S5). Then, the hydride transfer from [RuIIHX] to quinolinium salt A1 forms dihydroquinoline int-1 and its enamine tautomer int-2 along with regeneration of the catalyst precursor [RuIIX2]. Meanwhile, the condensation between aniline B1 and formaldehyde affords iminium B1'. Then, the β-nucleophilic addition of int-2 to B1' gives the β-aminoalkyl iminium int-3. Further, the electron-rich benzene ring of int-3 attacks the iminium motif from both the same (int-3b) and opposite (int-3a) sides of the H-atom at the β-site. In comparison, the form of int-3a (opposite side) is more favorable due to the less steric hindrance, thus affording product C1 with syn-selectivity after deprotonation of the coupling adduct int-4 (path a of Fig 7b, namely electrophilic aryl C−H aminoalkylation). Alternatively, the [4+2] cycloaddition of int-2 and B1' via endo or exo π-π stacking also rationalizes the formation of int-4 and product C1 (path b of Fig. 7b, via int-5 and int -4). Similarly, the generation of product D1 from isoquinoline is shown in Fig. 6c. The hydride transfer from [RuIIHX] to isoquinolinium salt A32 initially forms enamine int-6 (Scheme S3a, S3b and Fig. S6). Then, the β-capture of formaldehyde by int-6 followed by based-facilitated dehydration of int-7 and hydride transfer to alkenyl iminium salt int-8 forms β-methyl enamine int-9 (Scheme S3a and Fig. S7). Subsequently, the β-capture of B1’ by int-9 followed by intramolecular attack of the electron-rich phenyl ring to the iminium motif of int-10 from the sterically less hindered back side of the methyl group, or the [4+2] cycloaddition of int-9 and B1' via π-π stacking gives intermediate int-11. Finally, the deprotonation of int-11 generates product D1 with syn-diastereoselectivity (Fig 7c).
To better reveal the product formation including the unique cis-selectivity, computational study was preformed using the density functional theory (see details in SI). First, the participation of Mg(OMe)2, KOMe, and t-BuOK as the bases in the generation of [RuIIHCl] was calculated. The barrier for the transmetallation step with Mg(OMe)2 (14 kcal·mol-1, Fig. S10) is significantly higher than the other two bases (6.6 kcal·mol-1 for t-BuOK and 3.8 kcal·mol-1 for t-MeOK, Fig. S11 and Fig. 12) in the potential energy surfaces. This trend is in accordance with the fact that the higher charge of Mg2+ increases the stability of adduct E1 (Scheme 7a) and makes the dissociation of -MgOMe and transmetallation more difficult, thus leading to a slow forming rate of [RuIIHCl]. Correspondingly, a slow generation of enamine int-2 via hydride transfer from [RuIIHCl] to the azaarenium salt A1 is beneficial to the capture of int-2 by B1', and effectively avoids the formation of undesired N-benzyl tetrahydroquinoline A’’ (table 1).
Then, the potential energy profile computed for the conversion of int-2 and B1' to C1 is shown in Fig. 8a, and all of the structures were optimized in CH3OH solution. The formation of intermediate int-3 via β-nucleophilic addition of int-2 to B1' has an energy barrier of 14.0 kcal mol-1 (TS4), which represents an exergonic process, as int-3a is 6.3 kcal·mol-1 higher than int-2. In comparison, the formation of int-3b has a similar energy barrier of 14.5 kcal mol-1 (TS4'), the torsion of C3-N2 bond of int-3a to form int-3b has a barrier of 8.7 kcal mol-1 (TS5). However, the attack of the aniline benzene ring to the iminium motif of int-3b from the same side of the pyridyl β-H is less favored, which is due to the high stereoscopic hindrance of the N-ethyl group and pyridyl β-H as well as the long distance (~5.7 Å) between the pyridyl α-carbon (C1) and aniline ortho-carbon (C5). Thus, int-3a becomes a favorable intermediate. Starting with int-3a, the attack of aryl C5-atom on the C1-atom to form int-4 with a barrier of 15.5 kcal mol-1 (TS6) represents an exergonic reaction, as int-4 is 10.8 kcal mol-1 higher than int-3a. Finally, the formation of product C1, via deprotonative aromatization of the coupling adduct int-4 has no energy barrier, is favored from a thermodynamic point of view (ΔG = -54.1 kcal mol-1). In terms of the [4+2] cycloaddition of B1' and int-2, the manner of endo π-π stacking encountered commonly in the Diels–Alder reactions has a significant energy barrier of 39.0 kcal mol-1 (TS7). So, this pathway is disfavored. Meanwhile, we also found that it is difficult to form the transition state of exo π-π stacking due to the higher steric hindrance and long interaction distance. Based on the computational studies, path a shown in Fig. 7b is believed to be a favorable way in generating product C1.
As for the formation of requisite intermediate int-9, it involves four main steps (Fig. 8b): the β-addition of int-6 to HCHO (int-6 → int-6'), proton transfer from the methanol (int-6' → int-7), Mg(OMe)2-induced proton abstraction and dissociation of OH- (int-7 → int-7' → int-8), and hydride transfer (int-8 → int-9). Noteworthy, the formation of int-8 clearly proceeds under the assistance of Mg2+, and the hydride transfer (int-8 → int-9) by the [RuHCl] complex requires to overcome an energy barrier of 11.6 kcal·mol-1 (TS11), and the reaction is endothermic by 6.1 kcal mol-1. In comparison with this process, other parts can easily take place with a maximum barrier of 8.8 kcal mol-1 (TS8). Once int-9 has formed, the formation of D1 from int-9 undergoes a similar way of C1 generation from int-3 (Fig. 8a): β-addition of int-9 to B1', intramolecular cyclization via C1-C5 bond formation, and based-promoted deprotonation to yield product D1. The calculations show that the steps from int-9 to D1 have a slightly higher barrier than the corresponding transition state from int-3 to C1 (17.5 kcal·mol-1 for TS13 vs 15.5 kcal·mol-1 for TS6).