Modular Access to Azepines by Directed Carbonylative C-C Bond Activation of Aminocyclopropanes.

A modular Rh-catalyzed entry to azepines is outlined. Under a CO atmosphere, protecting group directed C-C bond activation of aminocyclopropanes provides rhodacyclopentanones. These intermediates are effective for intramolecular C-H metalation of either an N-aryl or N-vinyl unit en route to azepine ring systems. Thus, byproduct-free heterocyclizations are enabled by sequential C-C activation and C-H functionalization steps.

The azepine ring system is present in a wide range of bioactive compounds, with this motif or closely related variants representing the core structure of over 25 pharmaceutical products.[1] For example, the drugs Tolvaptan[2] and Benazepril,[3] and the beef improvement agent Zilpaterol,[4] all contain the tetrahydrobenzo[b]azepine ring system (Scheme A). Despite their established pharmaceutical value, azepines and other larger (>6-membered) N-heterocyclic rings are underrepresented in drug discovery libraries,[5] primarily because of a lack of efficient and modular routes for their preparation. Accordingly, synthetic methods that can address this issue are in high demand.

Scheme 1

Recently, as part of a broader program,[6] we outlined an N-heterocyclization strategy where “capture” of transient rhodacyclopentanones 2 by tethered nucleophiles occurs in advance of a C–Nu bond forming “collapse” to the targets (Scheme B).[6e] Conceptually, this approach is appealing because it harnesses the strain embodied within readily available and stereodefined aminocyclopropanes 1 for reaction initiation, and achieves otherwise challenging ring closures via the intermediacy of kinetically accessible bicycles 3. In our proof-of-concept studies, ureas were used as the nucleophile (1, R3 = NHR), such that intermediates 2 were converted to 1,3-diazepanes by C–N reductive elimination from 3.[6e] Outside of π-insertion processes, the reactivity of rhodacyclopentanones is relatively unexplored,[7−9] rendering further extension of our approach uncertain. However, it is well-established that Rh(III)-complexes can promote aryl C–H metalation in other contexts,[10] and we considered whether this type of process might be exploited from 2 to provide benzazepines 6. Specifically, for cyclopropanes 1 where R2 = aryl, we envisaged accessing targets 6 by a sequence of carbonyl directed C–C oxidative addition, metallacyclobutane carbonylation (to 2), aryl C–H metalation (to 5), C–C reductive elimination, and protodemetalation. At the outset, the viability of this design was considered tentative because of the absence of reports where rhodacyclopentanones engage in C(sp2)–H metalation or C(sp2)–C(sp2) reductive elimination processes. Nevertheless, outlined below is the successful realization of this approach, which provides striking examples of how metal-catalyzed carbonylative redistribution of C–C and C–H bonds can be harnessed in reaction design.

Our investigations commenced by exploring the carbonylative cyclization of 1a, which was readily prepared in two steps via N-arylation of cyclopropylamine and subsequent Cbz-protection (see the Supporting Information (SI)). Here we found that, using 1 atm CO, the combination of [Rh(cod)2]OTf (7.5 mol %), P(4-CF3C6H4)3 (15 mol %), Na2SO4 (30 mol %), and 2-NO2C6H4CO2H (100 mol %) enabled the generation of 6a in 82% isolated yield. Notably, neutral Rh-precatalysts were completely ineffective and electron-neutral/-rich P-based ligands provided only traces of the product (see the SI). 6a was formed in less than 30% yield in the absence of 2-NO2C6H4CO2H, with this additive likely acting as a proton reservoir to facilitate the final protodemetalation step, which otherwise would rely solely on the proton released during the conversion of 2 to 5.

With optimized conditions in hand, we explored the scope of the process with respect to the aromatic unit (Table ). A broad range of arenes are tolerated, with electron-rich (e.g., 1f), electron-poor (e.g., 1d), and heteroaromatic systems (e.g., 1j) all participating with similar levels of efficiency. For systems with meta-substituents (1g and 1i), C–C bond formation is highly regioselective, occurring at the more sterically accessible ortho-position. In the case of 1h, where both meta-positions are substituted, cyclization to 6h could be achieved, but the yield of the process was lower than for other examples. At the present stage, the protocol does not tolerate acyclic ortho-substituents. The method exhibits useful levels of flexibility with respect to the directing group. For example, other classes of carbamate can be used, as demonstrated by Fmoc-directed cyclization of 1b which provided 6b in 79% yield. Cyclic amides can also direct the process, and this allowed access to tricyclic systems 6k and 6l in 66% and 62% yield, respectively. However, acyclic amides and N-urea-based directing groups are ineffective (see the SI), highlighting the finely balanced nature of the process, where the DG must be Lewis basic enough to promote C–C bond activation, but also sufficiently labile to allow aryl C–H metalation (2 to 5).[11]

Table 1

Scope of the Aromatic Component

Reaction carried out at 140 °C.

At this stage, a series of experiments was undertaken to probe the mechanism of the process, and this insight proved fundamental in guiding extensions of the approach outlined later (Scheme ). When a 1:1 mixture of 1f and 1c was exposed to standard conditions, an approximately equimolar ratio of 6f and 6c was obtained at low conversion (eq 1). A similar competition experiment involving a 1:1 mixture of 1a and deuterio-1a revealed a relatively small kinetic isotope effect (kH/kD = 1.44, eq 2).[12] Taken together, these experiments suggest that aryl C–H metalation is not turnover limiting. Further insight was gained by running the carbonylative cyclization of 1a in the presence of CD3OD (300 mol %). At full conversion (eq 3), deuterium incorporation in the product deuterio-6a′ was observed at C-2 (12%), C-4 (23%), C-6 (9%), and C-9 (15%). Deuteration at C-2 is consistent with protodemetalation at this position (to 6) after C–C reductive elimination from 5. At partial conversion (eq 4), lower levels of C-6 deuteration were observed (deuterio-6a″), which is consistent with exchange at this position occurring via ketone directed C–H activation of the product. In support of this, exposure of product 6a to standard catalytic conditions in the presence of CD3OD resulted in 12% deuterium incorporation at C6 (eq 5). In this experiment, significant exchange also occurred at C-4 (28%), suggesting that deuterium incorporation observed at this position in eqs 3 and 4 arises via enolization of the product. Accordingly, the key issue remains as to how deuterium incorporation at C-9 of deuterio-6a′/6a″ occurs? Equation 5 indicates that this is not via C–H activation of the product 6a. Analysis of recovered starting material in eq 4 revealed approximately 15% deuterium incorporation at the ortho-positions (deuterio-1a′); however, a similar exchange experiment involving 7, which lacks a cyclopropyl unit, resulted in no deuterium incorporation at the ortho-positions (eq 6). This seemingly rules out an exchange pathway involving carbonyl directed C–H activation of 1a. Overall, the collective observations provided in eqs 3–6 indicate that deuterium exchange at C-9 of deuterio-6a′/6a″ and at the ortho-C–H bonds of deuterio-1a′ does not occur directly from 1a or 6a, but instead via reversible formation of another catalytically generated intermediate, which is most likely bicycle 5.[13] Indeed, our previous studies have shown that rhodacyclopentanone formation (1 to 2) is highly reversible using cationic Rh-systems,[6b,6c,6e] and the current experiments extend this reversibility to the aryl C–H metalation step. Accordingly, for the current process, we favor C–C reductive elimination as the first irreversible step, and further evidence supporting this assertion is given later.[14]

Scheme 2

Mechanistic Experiments

Substituted aminocyclopropanes can be accessed by Curtius rearrangement of cyclopropyl carboxylic acids, and these, in turn, can be synthesized easily in enantioenriched form.[15] Consequently, the reaction design outlined in Scheme B potentially allows access to benzazepines bearing stereodefined substituents that might be challenging to install using conventional methods. Key requirements for 1,2-disubstituted cyclopropane-based processes include (a) regioselective generation of the rhodacyclopentanone intermediate (2) and (b) transfer of this regiochemistry to the product (vide infra). Previous studies have revealed that directed rhodacyclopentanone formation from trans-1,2-disubstituted cyclopropanes occurs with high selectivity via the less hindered proximal C–C bond (bond a in 1m–o, Scheme A).[6c] However, as outlined above, the formation of 2 is likely reversible, such that product regiochemistry can be subject to Curtin–Hammett selectivity.[16] Indeed, initial attempts to access benzazepine 6m (R = Me) using the conditions from Table afforded approximately equal ratios of C-3 and C-4 adducts 6m and 6m′. The former arises from the kinetically favored rhodacyclopentanone 5m; however, subsequent C–C reductive elimination is likely to be slow because it requires the bulky Rh-center to move into the proximity of the C-3 methyl substituent. As such, reversible metallacycle formation provides kinetically disfavored regioisomer 5m′, which undergoes more facile C–C reductive elimination to provide C-4 methylated regioisomer 6m′. Electron-deficient ligand systems are known to accelerate C–C reductive elimination,[17] and so we examined whether replacement of P(4-CF3C6H4)3 with poorer donors might enforce selective generation of 6m from 1m (Scheme B). These studies revealed that 6m could be generated with essentially complete selectivity over 6m′ using either As(4-CNC6H4)3 or P(C6F5)3 as ligand. The latter offered the greatest efficiencies with 6m isolated in 68% yield and >99:1 e.r. from enantiopure 1m. These new conditions extended to more sterically demanding systems 1n and 1o, which cyclized to provide 6n and 6o as single regioisomers in 52% and 67% yield, respectively (Scheme C); 6o was prepared from racemic 1o, whereas 6n (99:1 e.r.) was accessed from enantioenriched 1n (99:1 e.r.). For the processes in Scheme C, 3-CNC6H4CO2H offered higher efficiencies than 2-NO2C6H4CO2H, which was used earlier (see Table ). So far, we have been unable to obtain high selectivity for the formation of C-4 substituted products (6m′–o′); however, substitution at this position can be introduced easily using enolate chemistry.[18] For example, highly diastereoselective conversion of 6m to α-functionalized products 8a and 8b was readily achieved, which demonstrates the feasibility of introducing C–C or C–heteroatom bonds at C-4.

Scheme 3

Benzazepines via trans-1,2-Disubstituted Cyclopropanes

The heterocyclizations described so far use an N-aryl unit as the nucleophilic component, and we questioned whether further classes of process might be achieved using other types of π-nucleophile. Specifically, we hoped to access nonbenzofused azepines (6p–t) by harnessing N-vinyl nucleophiles (1p–t) (Table ). In the event, by adapting of our previously developed conditions, carbonylative heterocyclization of 1p to 6p and 1q to 6q occurred in 61% yield and 42% yield, respectively. Here, the carboxylic acid additive was omitted as it promoted competitive polymerization of 1p and 1q. However, for subsequent examples (6r–t), an acid additive was beneficial, with the exact choice made on a case-by-case basis. A key aspect of the method in Table is that the substrates (1p–t) can be accessed in one pot from the corresponding ketone (see the SI), which, in turn, allows the two-step synthesis of a wide variety of interesting and challenging heterocyclic ring systems. From a mechanistic viewpoint, these processes are significant; in principle, the carboxylic acid additive could facilitate CMD-type metalation of the C(sp2)–H bond (2 to 5),[19] but the successful synthesis of 6p and 6q in the absence of acid suggests that this is not the case. Accordingly, we suggest that, for both N-aryl and N-vinyl systems, C–H metalation occurs by nucleophilic attack of the electron-rich π-system onto the Rh-center of 2, and that the primary role of the acid additive is to facilitate the final protodemetalation step.

Table 2

Processes Using N-Vinyl Carbamates

To conclude, we outline processes where rhodacyclopentanones generated by directed carbonylative C–C bond activation are captured by C-based nucleophiles en route to benzazepines and nonbenzofused variants. Ring systems of this type are difficult to construct in a modular fashion using conventional approaches,[20] and the methodology addresses this issue in an atom and step economical manner. Our reaction design is based on the exploitation of rhodacyclopentanones for C(sp2)–H metalation and subsequent C(sp2)–C(sp2) bond forming reductive elimination.[9,14] One can easily envisage harnessing these fundamental mechanistic steps in further rhodacyclopentanone-based methodologies. Studies toward this broad goal are ongoing and will be reported in due course.