Carbonylative N-Heterocyclization via Nitrogen-Directed C-C Bond Activation of Nonactivated Cyclopropanes.

Under Rh-catalyzed conditions, secondary amines and anilines function as directing groups to facilitate regioselective C-C bond activation of nonactivated cyclopropanes. The resulting amino-stabilized rhodacycles undergo carbonylative C-N bond formation en route to challenging seven- and eight-membered lactams. The processes represent rare examples where C-C bond oxidative addition of nonactivated cyclopropanes is exploited in reaction design.

Processes enabled by the oxidative addition of C–C bonds to transition metals (termed here as “C–C bond activation”) are of emerging strategic importance.[1] Predominant methodologies harness strained carbocycles to facilitate the C–C cleavage process, such that cyclopropane derivatives are commonly employed as initiating motifs (Scheme A).[2−5] For example, C–C bond activations of alkylidene cyclopropanes,[2] cyclopropenes,[3] and vinyl cyclopropanes have found widespread use;[4,5] in these cases, the fused, cyclic, or adjacent π-unsaturation predisposes the system to metal insertion. By contrast, the use of simple and more readily available nonactivated cyclopropanes is much rarer and is limited to just a handful of processes[6] outside of simple reduction or isomerization reactions.[7] Key issues include (a) the more challenging C–C bond activation process, (b) achieving regiocontrol, and (c) the instability of the incipient metallacycle, which is prone to deleterious β-hydride elimination (Scheme A, dashed box).

Scheme 1

C–C Bond Activations of Cyclopropanes

We have previously reported a range of processes that are enabled by carbonyl-directed carbonylative C–C bond activation of aminocyclopropanes (Scheme B).[8−10] Here, the directing group accelerates the rate of C–C bond activation and controls regioselectivity, whereas fast carbonylation of the rhodacyclobutane provides relatively stable rhodacyclopentanones (Int-I). Aminocyclopropanes are electronically privileged systems for these processes, and extension of our weak directing group strategy to completely nonactivated cyclopropanes has proven highly challenging.[9d] Nevertheless, carbonylative C–C bond activations of nonactivated cyclopropanes are efficient in stoichiometric settings.[11] For example, McQuillin and Powell reported distinct regiochemical outcomes for the carbonylative insertion of [Rh(CO)2Cl]2 into benzyl- and phenyl-substituted cyclopropanes (Scheme C).[11c] To translate this reactivity into catalytic processes, we considered replacing the weak carbonyl-based directing group used in Scheme B with much stronger N-based variants (Scheme D). In this design, N-directed carbonylative C–C bond activation of 1 leads to more stable metallabicycles Int-II, which can then undergo C–N reductive elimination en route to medium ring lactams 2. In this report, we outline our efforts in this area, which now allow diverse nonactivated cyclopropanes to be employed as initiating motifs for challenging carbonylative heterocyclizations. In broader terms, this work demonstrates how amines can be used to direct efficient C–C bond activation processes,[12,13] thereby complementing recently reported amine-directed C–H functionalization reactions.[14]

Initial efforts to realize the processes depicted in Scheme D focused on accessing benzazepines by N-directed C–C bond activation of aryl-substituted cyclopropanes 1a–g (Scheme A). Here, NH-metalation was expected to trigger N-directed carbonylative C–C bond activation and provide 5,5-metallabicycles Int-IIa. These should then undergo C–N reductive elimination to give alkyl-Rh(I) species Int-IIIa, from which (reversible) β-hydride elimination generates C2–C3 unsaturated products 2a–g. Other mechanistic options leading to Int-IIIa are also potentially feasible.[15]N-Benzyl system 1a, which was prepared via cross-coupling with cyclopropylboronic acid, was exposed to various Rh catalysts under an atmospheric pressure of CO. Ultimately, it was found that the combination of [Rh(cod)2]BARF (5 mol %) and dimethyl fumarate (100 mol %) in PhCN at 120 °C was optimal (see the SI); under these conditions, unsaturated system 2a was isolated in 72% yield. The use of exogenous phosphine or arsine ligands[9d] or neutral Rh complexes (e.g., [Rh(cod)Cl]2) resulted in substantially lower efficiencies. PhCN, a coordinating solvent that has been shown to stabilize Rh systems in other contexts,[9c] was optimal, and less coordinating options, such as 1,2-dichlorobenzene, resulted in minimal conversion. The addition of dimethyl fumarate (100 mol %) was the most critical factor, leading to a cleaner reaction and an approximately 30% increase in yield. This additive may function as a π-bound ligand,[16] but, perhaps more importantly, it also serves as the oxidant for the conversion of 1a to 2a; this was confirmed by 1H NMR and GCMS analysis of the crude reaction mixture, which revealed the stoichiometric formation of dimethyl succinate (see the SI). This presumably arises via a hydrometalation–protodemetalation sequence (Scheme A), with the latter step mediated by the proton released upon conversion of 1a to Int-IIa.[15]

Scheme 2

Benzazepines by N-Directed Carbonylative C–C Bond Activation

BARF = tetrakis[3,5-bis(trifluoromethyl)phenyl]borate.

Using [Rh] (7.5 mol %), AsPh3 (15 mol %) as an additive, and mesitylene as the solvent. 16% of C2–C3 unsaturated 2g was also isolated (see the SI).

With optimized conditions in hand, the scope of the process was explored, focusing initially on the R1 group. Cyclizations of systems where R1 = aryl (1b,c) or H (1d) were similarly effective; these results show that the protocol tolerates an appreciably wide electronic and steric range for the directing N-center. The electronics of the aromatic unit also have minimal influence, such that cyclizations of electron-rich (1e to 2e) and electron-poor systems (1f to 2f) proceeded with similar levels of efficiency. Cyclization of pyridyl system 1g did not occur under optimized conditions, perhaps because of the coordinating ability of the pyridyl nitrogen. Eventually, we established modified conditions where dimethyl fumarate was replaced by AsPh3 (15 mol %) and mesitylene was used as solvent. These conditions delivered 2g′, which arises via protodemetalation of Int-IIIa, in 66% yield, alongside 16% yield of C2–C3 unsaturated system 2g.[17] The distinct oxidation level selectivity of this process is expected based on the absence of the dimethyl fumarate oxidant. Nevertheless, we have been unable to identify general conditions that allow a switch of product oxidation level.[18] Interestingly, an alternate approach to this aspect is to program selectivity using the relative stereochemistry of 1,2-disubstituted cyclopropanes (Scheme B). For cis-1h, N-directed rhodacycle formation and reductive elimination lead to Int-IIIb, where syn-β-hydride elimination can occur to provide unsaturated product 2h in 46% yield. Conversely, for trans-1h, reductive elimination generates diastereomeric intermediate Int-IIIc, where syn-β-hydride elimination is not possible, and so protodemetalation predominates to give saturated product 2h′.[15] Interestingly, for both cis- and trans-1h, oxidative addition occurred at the more sterically accessible proximal C–C bond b; the selectivity of the former contrasts other C–C bond activation processes that use cis-1,2-disubstituted cyclopropanes.[8]

The substrates employed so far possess a degree of activation because the C–C bond that is cleaved is benzylic. The results in Scheme build upon the stoichiometric studies outlined in Scheme C by offering unique examples where metallacycles derived from this insertion mode are integrated into productive catalysis.[11c] Although the N-based unit plays a key role in these processes, the wider utility of this approach for achieving C–C bond activations of completely nonactivated cyclopropanes was uncertain. C–C oxidative additions involving such systems are extremely difficult and represent a frontier research challenge of the field. To probe the viability of the current strategy, we selected cyclopropane 1i, where the electronics of the N-unit are similar to 1a (Scheme A). Remarkably, exposure of 1i to optimized conditions delivered azepine 2i in 88% yield, albeit as a mixture of regioisomers favoring C2–C3 unsaturation. Isomerization is presumably mediated by the Rh-hydride released from Int-IIId, and further optimization studies indicated that suppression of this is challenging. Accordingly, we developed a telescoped protocol wherein xantphos (20 mol %) and H2 (1 atm) were added at the completion of the heterocyclization process to promote alkene reduction (Scheme B).[19] This “in situ hydrogenation” protocol segues the Rh catalyst into a second productive process and enabled the isolation of saturated system 2i′ in 78% yield. Further studies demonstrated that these protocols can be used to convert a range of monosubstituted cyclopropanes 1j–p to either unsaturated systems 2j–p or saturated azepines 2j′–p′ with good levels of efficiency. Notably, bicyclic products can be accessed (2p and 2p′), and, consistent with the proposed N-directed mechanism (cf. 1i to Int-IId), bystander cyclopropanes remain intact (2o and 2o′).

Scheme 3

Azepines by N-Directed Carbonylative C–C Bond Activation

After the carbonylation first step, xantphos (20 mol %) was added, and the mixture was heated under H2 (1 atm, 120 °C, 24–48 h); in some cases, a solvent swap to xylene or mesitylene was necessary (see the SI).

Combined yield of a mixture of olefinic regioisomers (major isomer shown). Product ratios (C1–C2:C2–C3:C3–C4:C4-C5): 2j (1:6:3.7:1.3), 2l (0:8:1:0), 2o (1:4.5:0:0), 2p (1:1.7:0:0).

The reaction temperature was 130 °C.

2q was isolated in >99:1 er when enantioenriched 1q (>99:1 er) was used (see the SI).

4-NO2C6H4CO2H (20 mol %) was used as an additive.

Under slightly more forcing conditions, the “standard” protocol extended to trans-1,2-disubstituted cyclopropanes 1q–s, and these were converted, via oxidative addition of bond b, to cyclic enamides 2q–s with useful levels of efficiency (Scheme C). In these cases, alkene isomerization was not problematic because the alkyl Rh(I) intermediate Int-IIIe is forced to undergo β-hydride elimination via C4, and this enforced directionality facilitates equilibration to give the observed and thermodynamically favored C4–C5 unsaturation. For 2s, the non-benzylic C−C bond b of 1s underwent cleavage, which contrasts the studies in Scheme C and highlights how the N-directing group enforces distinct regioselectivity. Even very hindered cyclopropanes participate (Scheme D); for example, trisubstituted system 1t delivered unsaturated azepine 2t in 44% yield. Similarly, bicyclic cyclopropane 1u could be converted to the [4.3.1] ring of 2u′ in 61% yield, with the addition of 4-NO2C6H4CO2H proving beneficial in this case.[20] Here, β-hydride elimination is not feasible at the stage of Int-IIIf, and so protodemetalation occurs instead to give the saturated product.[15] These latter examples extend the state of the art significantly with respect to cyclopropane C–C bond activation scope. Notably, the starting materials are easily accessed by (directed) Simmons–Smith cyclopropanation of the corresponding alkene, and so the strategy appears to be well suited to target directed applications.[21]

A unifying feature of the processes described so far is the proposed intermediacy of N-stabilized 5,5-metallabicycles such as Int-IIe.[15] The templating effect these provide, coupled with the relief of cyclopropane ring strain during their formation, facilitates ring closures that might otherwise be challenging.[22] Accordingly, we sought to examine whether homologous intermediates (e.g., Int-IIg) might be harnessable because this would provide highly challenging eight-membered systems.[15] To this end, a variety of electronically and sterically distinct N-directing units were evaluated, leading to the finding that N-benzhydryl system 1v participates efficiently to provide target benzazocine 2v in 61% yield (Scheme A).[23] This was isolated as a single alkene regioisomer, favoring conjugation to the arene. The N-phenyl and N-benzyl analogues of 1v were significantly less effective, suggesting that a strong N-directing group is required, but that this must be sufficiently sterically shielded to prevent saturation of the Rh center via polycoordination. As with earlier studies, the protocol is relatively insensitive to the electronics of the arene, such that 2w and 2x were generated with similar levels of efficiency. This 6,5-metallabicycle approach also extends to completely nonactivated cyclopropanes (Scheme B). Aniline-directed heterocyclization of 1y provided 2y′ in 79% yield using the “in situ hydrogenation” protocol described earlier. This was necessary because the initial heterocyclization step generated 2y (not depicted) as a 7:2 mixture of C3–C4 vs C4–C5 alkene regioisomers. The telescoped heterocyclization–reduction protocol extended smoothly to the formation of 2z′ and 2aa′. In nondirected settings, alkyl-substituted cyclopropanes undergo preferential cleavage of the distal C–C bond;[7,9a] the proximal selectivities observed in Scheme and Scheme B offer strong support for an N-directed pathway.

Scheme 4

Azocines by N-Directed Carbonylative C–C Bond Activation

In summary, we outline a strategy that allows readily available nonactivated cyclopropanes to be harnessed efficiently in C–C bond activation processes. Our reaction design exploits secondary amines and anilines as strong directing groups for the carbonylative C–C bond insertion of Rh(I) systems. This allows the regiocontrolled generation of diverse amino-stabilized metallabicycles, which lead, via reductive elimination, to challenging seven- and eight-membered lactams. To the best of our knowledge, these studies encompass the first examples of catalytic processes where amines and anilines are used to direct C–C bond activation. Applications of this strong directing group strategy to other settings are currently being explored.