Catalytic Enantioselective Synthesis of γ-Lactams with β-Quaternary Centers via Merging C–C Activation and Sulfonyl Radical Migration

Abstract

Transition metal-catalyzed C–C activation has become synthetically valuable; however, it rarely involves single-electron downstream processes. To expand the repertoire of C–C activation, here we describe the discovery of a Rh-catalyzed enantioselective C–C activation involving migration of a sulfonyl radical. This reaction directly transforms cyclobutanones containing a sulfonamide-tethered 1,3-diene moiety into γ-Lactams with a β-quaternary center in excellent enantioselectivity. This unusual process involves cleavage of C–C and N–S bonds and subsequent formation of C–N and C–S bonds. The reaction also exhibits broad functional group tolerance and a good substrate scope. A combined experimental and computational mechanistic study suggests that the reaction goes through a Rh(I)-mediated oxidative addition into the cyclobutanone C–C bond, followed by a Rh(III)-triggered N–S bond homolysis and sulfonyl radical migration.

Graphical Abstract

Enabled by strategic reorganization of bond connections, transition metal (TM)-catalyzed activation of carbon-carbon σ-bonds (C–C) has emerged as a useful tool to access complex molecular scaffolds from readily available starting materials.¹ In particular, C–C activation via oxidative addition² allows metalation at both carbon terminus, where further functionalization can take place (Scheme 1a). To date, almost all the downstream transformations of the resulting TM–C bonds involve two-electron processes. For example, the insertion of an unsaturated unit into a cyclic substrate, namely a “cut-and-sew” protocol,³ provides a direct and atom-economical approach for multi-atom ring expansion. However, the merging of TM-catalyzed C–C activation⁴ with single-electron events remains largely unexplored, which, if successful, could open a new arena for deconstructive synthesis of biologically important compounds.⁵

Scheme 1.

Transition Metal-catalyzed C–C Activation

During our exploration of the “cut-and-sew” reaction between cyclobutanones^3,6 and 1,3-dienes, an unexpected C–C activation-enabled N–S homolysis was discovered when common sulfonamides were used as the linker (Scheme 1b). In this transformation, the C–C activation triggered an unusual sulfonyl radical migration and a C–N formation. Here, we describe our preliminary development of a Rh-catalyzed enantioselective synthesis of chiral γ-lactams from sulfonamide-tethered cyclobutanone/1,3-diene substrates. This method offers a catalytic asymmetric entry to functionalized chiral γ-lactams containing a β-quaternary center, which are commonly found in bioactive compounds⁷ (Scheme 1c) but non-trivial to prepare enantioselectively without stoichiometric chiral reactants.⁸

Sulfonamides are among one of the most robust structural motifs under typical reaction conditions;⁹ thus, they are commonly employed as linkers in cycloaddition reactions. When a sulfonamide-tethered 1,3-diene/cyclobutanone substrate (1a) was employed under the cationic Rh-catalyzed “cut-and-sew” conditions,¹⁰ it was surprising that the desired 5/6 bridged bicycle was not obtained. Instead, a γ-lactam with a N-4-sulfonyl-1,3-dienyl substituent (2a) was isolated as the dominant product,¹¹ whose structure was unambiguously conformed by X-ray crystallography. After further optimization, γ-lactam 2a was ultimately produced in 79% yield using Rh(COD)₂NTf₂ and dppbz as the catalyst with butyric acid as the additive (Table 1, entry 1).¹² In this transformation, the α C–C bond of the cyclobutanone and the N–S bond of the sulfonamide are cleaved, while a C–N bond is simultaneously formed with the sulfonyl group migrated to the terminal position of the 1,3-diene moiety. Nearly complete (1E, 3E) selectivity was observed at the 1,3-diene. Note that the 4-sulfonylbuta-1,3-dien-1-amine structure is very rare with almost no prior preparation.¹³

Table 1.

Selected Optimization of the Reaction with Cyclobutanone 1a

^aAll reactions were run on a 0.05 mmol scale for 17 h.

^bPyrazine was used as the internal standard to determine the yield by ¹H NMR.

^cIsolated yield on a 0.01 mmol scale. n.d. = not detected. dppbz = 1,2-bis(diphenylphosphino)benzene, Xantphos = 4,5-bis(diphenylphosphino)-9,9-dimethylxanthene.

Intrigued by this surprising transformation, a number of control experiments were conducted to gain some insights. First, both the Rh catalyst and the dppbz ligand were essential to this reaction (entries 2 and 3). Other phosphine ligands were much less efficient (entries 4 and 5). The reaction was not promoted by light as the “dark” condition still afforded comparable yield (entry 6). Among various carboxylic acid additives examined, butyric acid offered the highest yield possibly due to its higher boiling point and better solubility (for details, see Supporting Information). Much lower conversions were observed in the absence of butyric acid or using NaHCO₃ instead (entries 7 and 8). Using one equivalent of butyric acid gave slightly reduced yield (entry 9). Stronger tosylic acid resulted in a full conversion albeit with no desired product due to substrate decomposition (entry 10). Adjusting the reaction temperature was not beneficial, leading to less efficient reactions (entries 11 and 12).

Considering that an all-carbon quaternary stereocenter would be generated when changing the C3-methyl group in cyclobutanone 1a to other groups, a catalytic desymmetrization was next explored to achieve enantioselective synthesis of γ-lactams.¹⁴ With cyclobutanone 1b as the model substrate, to our delight, the desired γ-lactam 2b was ultimately obtained in 82% yield and 95% ee using MeO-BIPHEP (L2) as the chiral ligand and DMA as the solvent (Scheme 2).¹⁵ The 1:2 metal/ligand ratio appeared to be critical for both yield and reproducibility, though the reason remained unclear. DMA proved to be a better solvent for the enantioselective version of the reaction (for details, see Supporting Information). Among various chiral bidentate phosphine ligands surveyed, MeO-BIPHEP (L2) with a smaller dihedral angle afforded the highest yield and enantioselectivity. In general, ligands with axial chirality gave good to excellent enantiomeric excess (ee), and bulkier and electron-rich substituents on the phosphorus (such as L6 and L7) appear to hamper the enantioselectivity. In contrast, ligands with point chirality (e.g., L8 and L9) were less efficient.

Scheme 2.

Enantioselective Transformation of Cyclobutanone 1b.^a

^aAll reactions were run on a 0.05 mmol scale for 17 h. Yields were determined by ¹H NMR with pyrazine as the internal standard. ^bIsolated yield on a 0.01 mmol scale. ^cEnantiomeric excess (ee) was determined by chiral HPLC. DMA = N,N-dimethylacetamide.

With the optimized conditions in hand, the substrate scope for both the racemic and enantioselective reactions was investigated (Table 2). First, diverse substituents at the cyclobutanone C3 position can be tolerated. Various benzyl moieties with either electron-donating or -withdrawing groups can afford the desired γ-lactams in good yield and excellent enantioselectivity (2c-g). While the racemic condition gave a low yield for γ-lactam 2f that contains a nitrile group, 69% yield was nevertheless obtained with the chiral catalyst. For the 2-chlorobenzyl substituted substrate, while product 2h was obtained in moderate yield with the achiral catalyst, the asymmetric conditions resulted in a complex reaction mixture. It is likely that the ortho-chloro group is responsible for the undesired reactions under the more forcing asymmetric conditions. Besides methyl and benzyl groups, simple alkyls, such as ethyl (2i) and isopropyl (2j), as well as functionalized alkyl moieties (2k-2m), all delivered the desired γ-lactam products in 90–99% ee. Aside from sp³ substituents, cyclobutanones with a β-sp² group, such as phenyl (2n) and ester (2o), also proved to be competent substrates, again showing excellent enantioselectivity. A range of functional groups, including anisole (2d), fluoro (2e), chloro (2h), nitrile (2f), ester (2o), alkyl ether (2k), silyl ether (2l) and imide (2m), were compatible. In addition to tosyl (Ts), other sulfonyl groups, e.g., benzenesulfonyl (Bs, 2p) and p-nitrobenzenesulfonyl (p-Ns, 2q), can also undergo migration with excellent enantioselectivity albeit with a slightly compromised 3-E/Z selectivity at the diene; in contrast, mesyl (Ms) was much less efficient, giving <10% yield. The cyclobutanone bearing a hydrogen at the C3 position still afforded the desired γ-lactam (2r) but in moderate yield, likely due to the lack of Thorpe–Ingold effect for the desired cyclization. Finally, the more challenging one-carbon elongated substrate delivered the desired δ-lactam (2s), which shows promise to access six-membered-ring products. It is noteworthy that other nitrogen protecting groups, such as acetyl, did not migrate, and no other π-systems (alkenyl, allyl, aryl groups) except 1,3-diene¹⁶ can promote the same transformation under the current conditions.

Table 2.

Scope of Substrates^a

^aAll reactions were run on a 0.1 mmol scale for 17 h, isolated yield.

^b2-Me-THF instead of DMA.

^cRh(COD)₂SbF₆ instead of Rh(COD)₂NTf₂. N/A = not available.

To explore the synthetic utility, the reaction was first run on a larger scale, which gave even higher yield (Scheme 3a). In addition, the sulfonyl diene moiety can undergo several transformations (Scheme 3b). For example, the sulfonyl group can be rapidly removed by treatment with Mg and TMSCl in DMF at 0 °C for 1 h.¹⁷ An alternative approach involves a catalytic desulfonylative borylation followed by hydrolysis.¹⁸ The N-formyl lactam 9 can be prepared in excellent yield by ozonolysis. Moreover, the sulfonyl group can be replaced with a silyl moiety by a radical addition-elimination pathway (10),¹⁹ and the silyl group may serve as a handle for further functionalization. The diene moiety can be efficiently removed to give free lactam 11. Furthermore, hydrogenation of the double bonds provided alkyl sulfone 12 in nearly quantitative yield.

Scheme 3.

Scalability and Product Transformations

Finally, preliminary mechanistic studies were conducted to understand how the N–S bond was cleaved.²⁰ First, the crossover experiment using a 1:1 mixture of Bs-based substrate 1p and Ts-based substrate 1a yielded all the four products (2a, 2b, 2p and 2ap) (Scheme 4a). The formation of the crossover products (2b and 2ap) suggests that the sulfonyl transfer goes through an intermolecular process. It is likely that a free sulfonyl species is involved in the reaction and such species could be of either anionic, cationic, or radical character. Next, various control or trapping experiments were carried out to identify this species. The addition of exogenous sodium benzenesulfonate was found to inhibit the reaction and no Bs-incorporated product (2p) was observed, which could rule out the sulfonate anion-mediated pathway (Scheme 4b). On the other hand, attempts to capture any electrophilic sulfonyl species with electron-rich arenes were not fruitful, which is not consistent with the involvement of a cationic sulfonyl species. To examine the possibility of a sulfonyl radical-mediated pathway, phenyl benzenesulfinylsulfonate (compound 3, Scheme 4c), recently demonstrated by Bi and co-workers to be an excellent sulfonyl radical source,²¹ was used as the additive; indeed, product 2p containing the Bs group was formed together with the normal product (2b). While it is generally challenging to trap or detect free sulfonyl radical, the use of stoichiometric Rh(COD)₂NTf₂-dppbz complex allowed successful detection of the TEMPO-Ts adduct 4 by high-resolution mass spec. Finally, the C–C activation process appears to be critical for the sulfonyl migration, as the use of the analogous N-butyl substrate (5) gave no N–S bond cleavage with most of the reactant recovered (Scheme 4d).

Scheme 4.

Preliminary Mechanistic Studies

To gain a deeper understanding on how the N–S bond is cleaved, the DFT calculation was performed (Figure 1a). The computational results show that the C–C activation via oxidative addition has a moderate activation barrier with the cationic catalyst (see TS 1–1 versus TS 1–2), which is consistent with our prior mechanistic studies.^10b The resulting oxygen-coordinated five-membered rhodacycle (int 2) can be converted to a more stable nitrogen-coordinated int 3 through a barrierless ligand exchange, which contains a relatively longer N–S bond (1.83 Å vs 1.67 Å in int 2). From int 3, heterolytic bond cleavages to form int 4-2 and int 4-3 (Ts cations and anions) are highly endergonic (>54 kcal/mol). In contrast, the homolysis pathway is more thermodynamically favorable, which is due to formation of a stable Ts radical and a largely delocalized dienylamino radical, as depicted in the electron spin density image (Figure 1b). In addition, the flexible PES (potential energy surface) scan showed that the system energy changed with increasing the N–S distance; the smooth curve of the energy change suggested that no transition state likely exists in this homolytic cleavage process. After the N–S bond cleavage, the Ts radical attacks at the terminal (C4) carbon of the diene, which has both substantial spin density and SOMO orbital coefficient (see Supporting Information for details). An alternative Ts-radical-triggered Ts migration pathway has also been considered. The di-Ts-substituted radical species int4-0, generated via Ts radical addition into int 3, is more stable than int 4-1 by 6 kcal/mol, and the resulting ${\pi }_3^3$ system is expected to promote smooth homolysis of the N–S bond to regenerate Ts radical.²² This chain-reaction pathway could be initiated in the presence of sufficient concentration of Ts radical. Generated from either pathway, the 18-electron closed-shell intermediate int 5 then undergoes a proton-transfer or tautomerization process, facilitated by butyric acid or its conjugate base, to give int 6. The subsequent C–N bond reductive elimination proceeds through a low activation barrier (13.2 kcal/mol, TS 2, Figure 1c). The following alkyl-Rh intermediate int 7 undergoes protonation to release the product (2a) and to regenerate the substrate-catalyst complex (int 1-1). Note that the overall reaction process is thermodynamically favored by 30.1 kcal/mol.

Figure 1.

DFT Studies.^{23 a} Computations were performed with Gaussian 16 at level of B3LYP(d3)-LANL2DZ/6-31G(d,p) for geometry optimization, M06(d3) (SMD(THF))-SDD/6-311+G(d,p) for single point energy. ^b int 3a is a transition structure from int 3 to int 4-1 with a 2.48 Å N–S bond length. Its energy value shows as an orange dot in b-2. Spin density of int 3a is illustrated by Multwfn. ^c Structures of intermediates and transition states are illustrated by CYLview.

In summary, we have developed a Rh-catalyzed C–C activation-triggered sulfonyl radical migration reaction, which provides an efficient and enantioselective entry to access diverse γ-lactams that contain β-quaternary centers. The reaction shows broad functional group tolerance and a good substrate scope, which could be attractive for complex molecule synthesis. The unique mechanistic features discovered here, i.e., merging homolytic N–S bond cleavage with asymmetric C–C oxidative addition, could have broad implications beyond this work.