Introduction

Molecular rearrangements exist ubiquitously in modern synthetic chemistry, providing a powerful strategy to reorganize complex structures in an atom- and step-economic process through one-step chemical transformation1,2. Almost a century ago, the Beckmann rearrangement3,4, semi-pinacol rearrangement5,6,7, Smiles rearrangement8,9,10,11,12, Wolff rearrangement13,14,15,16,17, and others18 have been successively developed to allow for diversified group migration via parallel moving pattern (Fig. 1a). Meanwhile, the Cope and Claisen rearrangement19,20,21 and Witting rearrangement22,23 represent another type of cyclic transition state-based [3,3′]- and [2,3′]-sigmatropic shift, one of the most important features of these rearrangements involves synergetic C–X σ bond (X = C, O, N, etc.) formation and double-bond migration at pericyclic reaction-sites (Fig. 1b). To date, these classical group migratory and sigmatropic rearrangements have shown extremely potential in strategic synthesis of natural products, pharmaceuticals, and material molecules24,25,26. However, it is surprising that 1,2-sigmatropic rearrangement (1,2-STR) or 1,3-sigmatropic rearrangement (1,3-STR), which only involves two different reaction-sites, has remained unexplored. This protocol will probably establish an efficient platform to enable in-situ U-turn-like rearrangement, which refers to a U-shaped turn made by a molecular skeleton so as to head in the opposite direction from its original course.

Fig. 1: Strategies to access molecular skeleton rearrangement.
figure 1

a The traditional migratory patterns in Smiles, semi-pinacol, Beckmann, and Wolff rearrangements. b The sigmatropic shift in the Cope and Claisen rearrangement and the Witting rearrangement. c This work: Mn(I)-catalyzed 1,2-STR or 1,3-STR of α-aryl-β, γ-unsaturated alcohols.

Full size image

Although C–C σ bond activation belongs to a significantly challenging transformation due to the high C–C bond dissociation energy and kinetical inertia27,28, C–C σ bond functionalization has proven to be a straightforward approach to modifying complex carbon skeletons29,30,31,32,33,34,35,36,37,38,39,40. Unfortunately, the current unstrained C–C σ bond activation strategies generally lead to the loss of another carbon-containing component derived from C–C bond cleavage28,33. Thus, the development of unstrained C–C σ bond functionalization via rearrangement with high atomic and step-economy remains unmet challenges: (i) the efficient carbon–carbon activation-based group-switch generally requires 2–5 equivalents of coupling reagents36,37,39,40, but the stoichiometric ratio of the cleaved moieties and the parent carbon skeletons is only up to a maximum of 1; (ii) the cleaved and lost carbon-containing moieties generally possess high reactivity and suffer from rapid decompositions either by processes of nucleophilic attack or by oxidation35,36,37, leading to significant difficulty in their re-utilization. Nevertheless, given that ligand-directed C–C bond activation generally involves a cyclometalation process, which possibly once again traps the unsaturated carbonyl, vinyl or alkynyl species derived from C–C bond cleavage through the delicate balance between substrate-based reactivity and metal-based catalytical activity. Here, we show a Mn(I)-catalyzed carbon-skeleton reorganization of α-aryl-β, γ-unsaturated alcohols via C–C σ bond activation-based sigmatropic rearrangement (Fig. 1c).

Results

Investigation of reaction conditions

Earth-abundant first-row metal catalysts, which possess low cost, excellent sustainability, and environmentally benign properties, have attracted increasing attention in modern synthetic chemistry41,42,43,44,45,46. Nevertheless, the exploration of non-noble-transition metal-catalyzed carbon-skeleton rearrangements is very scarce47,48. The choice of Mn-catalysts was motivated by the notion that the valence electron configuration (3d54s2) of elemental manganese endows different oxidation states of manganese species with distinctive coordination to arenes, alkenes, and alkynes49,50. We therefore utilized 1-(1-(pyridin-2-yl)−1H-indol-2-yl)prop-2-en-1-ol (1a) as a model substrate to evaluate the feasibility of Mn-catalyzed [1,3]-sigmatropic rearrangement via carbon–carbon sigma bond activation (Table 1). First, various manganese salts such as Mn(OTf)2, Mn2(CO)10, and Mn(CO)5Br, etc. were investigated in toluene to optimize the catalyst system (entries 1-5), we were quite pleased to find that the treatment of substrate 1a with Mn2(CO)10 and Mn(CO)5Br at room temperature (25 oC) did furnish 23% and 27% yield of 3-(1H-indol-2-yl)propanal 3a (entries 4 and 5), respectively, in which the allylic-alcohol moiety of 1a underwent a 1,3-STR through an in-situ carbon-skeleton-rearrangement; Meanwhile, 2-unsubstituted indole 2a, which derived from the carbon–carbon bond cleavage and protonation, could also be obtained in 3% yield (entry 5). To our delight, after various reaction temperatures were scouted by employing Mn(CO)5Br as catalysts (entries 5–8), it was found that running this transformation at 75 oC brought us increased conversion, affording 61% yield of 1,3-STR product 3a and 7% yield of 2a (entries 5–8 vs. 7). More satisfactorily, screening solvent systems further confirmed that 1,2-dichloroethane (DCE) showed a positive effect on this reaction, significantly increasing the reaction yield to 80% with excellent chemoselectivity without the formation of 2a (entries 7–12 vs. 13). Notably, Cp*Co(CO)I2 and Cp*Rh(CH3CN)3(SbF6)2 catalysts surveyed under the optimized conditions delivered 3a with very poor reaction conversions and chemoselectivity (entries 14 and 15).

Table 1 Reaction developmenta
Full size table

Substrate scope

With established reaction conditions, the α-(2-indolyl)-substituted secondary allylic alcohols were evaluated by using the optimized conditions. As summarized in Fig. 2, comparison with electroneutral indolyl-substituted allylic-alcohol 1b, Mn(I)-catalyzed [1,3]-sigmatropic rearrangement of α-[2-(3′-alkylindolyl)] and α-[2-(3′-aryllindolyl)]-substituted allylic alcohols including 1a and 1c-1e could efficiently produce 2-carbonylethylindoles 3a-3e in good to excellent yields (66–80%), regardless of the steric hindrance from C3-substituents of indolyl ring (entry 1). Meanwhile, electron-donating 5-Me, 5-MeO, 5-halo-, and even strong electron-withdrawing 6-CF3-indole-based allylic alcohols 1f-1j in which indolyl C3-position kept unsubstituted, could also successively undergo carbon–carbon activation and 1,3-STR to provide 3f-3j in 65–73% yields (entry 2). Moreover, N-electron-rich and N-electron-deficient pyridyl-substituted α-(2-indolyl)alcohols (1k-1o) were also well-tolerable in this transformation to assemble β-(2-indolyl)propanals 3k-3o with good conversions (66-70%) (entry 3). Again, switching terminal allylic alcohols to the internal allylic-alcohol 1p could still furnish the corresponding rearrangement product 3p in 62% (entry 4). Of noted, α-(N-(2-pyrimidyl)indolyl) allylic-alcohol 1q and α-(N-(2-pyridyl)pyrrolyl) allylic-alcohol 1r also participated in this carbon-skeleton rearrangement, providing 3q (76%) and 3r (35%), respectively (entry 5).

Fig. 2: Secondary allylic-alcohol scope.
figure 2

All the reactions were performed in a sealed tube, followed by flash chromatography on SiO2. The yields were determined by isolation.

Full size image

In comparison with the photocatalyzed 1,3-alkyl shift of tertiary allylic alcohols reported by Knwoles51, our catalysis system could be applied to α-aryl-tertiary allylic alcohols 4, which underwent chemoselective 1,3-aryl transposition (Fig. 3). Among them, α-(2-indolyl)-α-alkyl-allylic alcohols (4a-4c) and α-(2-indolyl)-α-aryl-allylic alcohols (4d-4f) worked well to deliver 2-indolyl-tethered ketones 5a-5f in 65–85% yields. Moreover, switching α-(2-indolyl)- allylic alcohols to α-phenyl-allylic alcohols (4g-4k) in which pyrazole was utilized as a directing group, could also produce the corresponding 2-phenyl-tethered ketones 5g-5k in excellent reaction conversions (71-86%). Importantly, the potential breadth of the utility of this methodology is further illustrated with different six- and seven-membered α-vinyl cycloalkylalcohols 4l and 4m, which are effective substrates for assembling sixteen- and eighteen-membered macrocyclic molecules 5l (40%) and 5 m (62%, CCDC 2121152) via a carbon–carbon σ bond cleavage-based bimolecular [2n + 4] coupling-cyclization (see Supplementary Fig. 7 for the possible reaction mechanism of the formation of 5l and 5m).

Fig. 3: Tertiary allylic-alcohol scope.
figure 3

All the reactions were performed in a sealed tube, followed by flash chromatography on SiO2. The yields were determined by isolation. aThe reaction temperature is 100 oC.

Full size image

Encouraged by Rh(I)-catalyzed 1,3-alkynyl shift of alkynyl alkenyl carbinols in which Csp3-Csp bond cleavage occurred52, we further evaluated the rearrangement reactivity of different α-(2-indolyl) propargyl alcohols (Fig. 4). Interestingly, Mn(I)-catalyzed 1,2-STR of internal propargyl alcohols (6a-6f) could occur to produce α-indolyl-α-vinyl aldehydes 7a-7f in 55-84% yields, and the carbon-skeleton rearrangement model belongs to 1,2-STR instead of 1,3-STR. Among them, the exact structure of 7f (CCDC 2121154) was determined by its single-crystal X-ray diffraction. However, if terminal propargyl alcohol 6g was subjected to the same reaction system, the desired 1,3-STR product 3-(2-indolyl)-propargyl aldehyde 7g (38%) was produced possibly due to the absence of steric hindrance from the terminal substituent of propargyl alcohols. In comparison, switching terminal propargyl alcohol 6g to the substrate 6h with a larger size of methyl group at the alkyne terminus produced both 1,2-STR product 7h (35%) and 1,3-STR product 7h′ (62%, E/Z = 1:3.3)53. Again, we explored the reactivity of α-(2-indolyl)-α-alkyl-propargyl alcohols (6i and 6j) and α-(2-indolyl)-α-aryl-propargyl alcohol 6k, and found these tertiary alcohols still smoothly underwent regioselective 1,2-STR to provide the corresponding α-(2-indolyl)-β-alkyl-α, β-unsaturated ketones 7i (66%), 7j (48%) and α-(2-indolyl)-β-phenyl-α,β-unsaturated ketone 7k (77%), respectively. Similarly, 1-(2-indolyl)−3-phenylprogargyl alcohols with different substituents in the indolyl C-5 position were also amenable to this transformation, producing the desired α-(2-indolyl)-allylic aldehydes 7l–7q in 38–60% yields, in which the substituents including methyl- (6m), methoxyl- (6n), fluoro- (6o), bromo- (6p), trifluoromethyl- (6q) showed an apparently electronic effect on the carbon-skeleton rearrangement. More complex alkoxycarbonylphenyl-tethered propargyl alcohols, which were derived from the natural menthol (6r) and stigmasterol (6s), could still undergo intramolecular 1,2-STR to deliver the corresponding indole-containing menthol and stigmasterol derivatives 7r (73%) and 7s (64%), respectively. Importantly, it should be pointed out that the Mn(I)-catalyzed 1,2-STR of α-alkynyl cycloalkylalcohols 6t and 6u led to a monomolecular [n + 1] ring-extension instead of bimolecular [2n + 4] coupling-cyclization, furnishing seven-membered 2-benzylidenecycloheptan-1-one 7t (40%, CCDC 2133328) and eight-membered 2-benzylidenecyclooctan-1-one 7u (34%), respectively.

Fig. 4: Propargylalcohol scope.
figure 4

All were performed in a sealed tube, followed by flash chromatography on SiO2. The yields were determined by isolation. aThe reaction time is 24 h. bThe reaction time is 72 h. cDiastereomeric ratio (d.r.) was measured by 1H NMR spectrum.

Full size image

Application

To showcase our system’s synthetic utility (Fig. 5), the pyrimidyl moiety from aldehydes 3q could be readily removed to access free (N-H) β-(2-indolyl)ethyl-1,3-dioxolane 8 (75%). Moreover, the indole-derived aldehydes 3a and 7a could be further utilized as versatile platform-molecules for diversity-oriented assembly of 4-(2-indolyl)alkyne 9 (71%), 4-(2-indolyl)−5-phenylpenta-2,4-dienenitrile 10 (37%) and 2-(2-indolyl)−1,4-diphenylbuta-1,3-diene 11 (67%) by coupling with α-diazo-phosphonate, acetonitrile and diethyl benzylphosphonate, respectively. Of course, α, β-unsaturated aldehyde 7a could be regioselectively reduced by NaBH4 to give 2-(2-indolyl)−3-phenyl-allylic alcohol 13 (82%). Interestingly, the reaction between aldehyde 7a and Me3SiOK still provided unexpected 2-phenethyl-indole 12 (44%) through reductive deformylation. Finally, 4-(2-indolyl)pyrazole 14 (42%) was obtained after subjecting 7a to the coupling-cyclization with phenylhydrazine.

Fig. 5: Synthetic applications.
figure 5

a Removal of pyrimidyl group. b Alkynylayion of aldehydes. c Cyanation of aldehydes. d Arylvinylation of aldehydes. e Reduction of carbon–carbon double bonds and carbonyl group from α, β-unsaturated aldehydes. f Reduction of aldehydes. g Coupling-cyclization of α, β-unsaturated aldehydes with hydrazines.

Full size image

Mechanistic investigations

To gain insights into the mechanism of the Mn(I)-catalyzed sigmatropic rearrangement, we performed the carbon-skeleton rearrangement of 1a with D2O under the standard reaction system, and obtained d-2a (77%) in which 45% D was incorporated into the α-position of aldehyde (Fig. 6a), indicating that either keto-enol isomerization or protonation of carbon-metal bonds was involved in this transformation (the deuteration of the product 3a was also performed in the presence of D2O (5.0 equiv.) under the standard conditions, and 7.5% D was observed at the α-position of aldehyde 3a, please see SI for more details). Evaluating the reactivity of α-(2-(N-phenylindolyl)allyilic alcohol 1u as a substrate only gave an intramolecular cyclization product 3u′ (61%) instead of 1,3-STR product 3u (Fig. 6b), demonstrating that pyridine played a key chelation-assisted role to enable this carbon-skeleton rearrangement.

Fig. 6: Preliminary mechanism studies.
figure 6

a H/D exchange experiment. b Intramolecular cyclization of α-[2-(N-phenyl) indolyl]allylic-alcohol. c Csp2-Csp3 σ bond activation-directed carbonylethylation of indoles. d Csp2-H bond activation-directed carbonylethylation of indoles. e Mn(I)-catalyzed cross-coupling reaction via 1,3-STR.

Full size image

Comparison with the reactivity of allylic-alcohol 1a and 2-unsubstituted N-pyridylindole 7 under the same reaction conditions (Fig. 6c vs. 6d), it was found that Mn(I)-catalyzed Csp2-H bond cross-coupling of 7 with acrylaldehyde at 45 oC could only give 3a with poor conversion (75% vs. 15%). These observations pointed to an understanding mechanism that 2-unsubstituted indole 7 was not a possible reaction intermediate derived from C–C σ bond cleavage of 1a; Meanwhile, the subsequent Mn(I)-catalyzed C-H activation and cross-coupling with acrylaldehyde were not possibly involved in this rearrangement. The intermolecular competing reactions between 4a and 1g produced 1,3-STR products 5a (64%), 5n (31%), 3c (43%) and 3a (24%), indicating that cyclomanganated species and the in-situ generated acrylaldehydes probably underwent different site-selective Mn-carbonyl and Mn-alkene complexation via a step-wise process (Fig. 6e).

Based on these control experiments, the possible reaction mechanism is proposed in Fig. 7. The interaction between allylic-alcohol 1a with Mn(CO)5Br produced cyclomanganated species A, followed by chelation-assisted β-aryl elimination to give six-coordination Mn-carbonyl complexes B, in which the “carbonyl oxygen” of the in-situ generated acrylaldehyde coordinated to Mn(I) cations. Subsequently, Mn-aldehyde intermediate B underwent an intramolecular ligand exchange to give Mn-alkene complexes C in Path a. The subsequent migratory insertion and protonation of carbon-Mn bonds of complexes C afforded the 1,3-STR product 3a. Of course, an alternative intramolecular Michael-addition initiated by Mn-aldehyde intermediate B via Path b could possibly proceed to furnish enol oxygen-coordinated Mn(I)-complexes E, which then underwent a cascade protonation and keto-enol tautomerization to afford 3a. Instead, indole 7 and Mn-complexes F as possible intermediates, were not involved in this [1,3]-sigmatropic rearrangement, supported by our experimental exclusion of the C-H activation (Fig. 6d).

Fig. 7: Proposed reaction mechanism.
figure 7

Path a Mn(I)-catalyzed 1,3-STR via intramolecular ligand exchange and migratory insertion. Path b Mn(I)-catalyzed 1,3-STR via intramolecular Michael-addition.

Full size image

Density functional theory (DFT) studies were performed to explore the Mn(I)-catalyzed carbon-skeleton rearrangement (the potential free energy profiles of the major reaction pathways in Fig. 8 and the full version in Supplementary Fig. 8). The transformation begins with the cyclomanganated complex A. Both the cisoid and the transoid isomers of the formed enal are considered for the β-aryl elimination. The Gibbs free energy of transoid-A is 2.5 kcal/mol higher than that of cisoid-A due to the steric effect between the vinyl and the indole fragments. The release of one CO molecule from the cyclomanganated complex A gives intermediate A1 (cisoid-A1: 9.3 kcal/mol vs. transoid-A1: 10.6 kcal/mol), providing a vacant site for the β-aryl elimination. The chelation-assisted β-aryl elimination from cisoid-A1 or transoid-A1 then goes through transition states cisoid-TS1 (16.2 kcal/mol) or transoid-TS1 (18.5 kcal/mol), respectively, to furnish the six-coordination Mn-carbonyl intermediates B (cisoid-B: 1.4 kcal/mol vs. transoid-B: −1.2 kcal/mol). The Gibbs free energy of activation of cisoid-TS1 is slightly lower than that of transoid-TS1 by 2.3 kcal/mol due to the similar steric effect. After C-C sigma bond cleavage, as expected the transoidal enal transoid-B (−1.2 kcal/mol) is more stable than the cisoidal isomer cisoid-B (1.4 kcal/mol). The Mn-carbonyl intermediates B could isomerize to the Mn-alkene intermediates C (cisoid-C: 4.8 kcal/mol vs. transoid-C: 2.7 kcal/mol). Subsequently, the Mn-alkene intermediate C initiates the migratory C = C insertion to the afford cyclo Mn-alkyl intermediate D through the four-membered ring transition state TS2 (Path a). Compared with cisoid-TS2 (11.4 kcal/mol), transoidal C isomer results in a more stable migration insertion transition state transoid-TS2 (10.3 kcal/mol). Nevertheless, the formed Mn-alkyl intermediate D will recombine the CO ligand and release product 3a after proton transfer and keto-enol tautomerism. The OH of 1a coordinates with the metal center of intermediate DCO to promote the intramolecular proton transfer from 1a to acrylaldehyde via transition states cisoid-TS4 (4.2 kcal/mol) or transoid-TS4 (3.4 kcal/mol), which then releases 3a after keto-enol tautomerism. On the other hand, the Mn-carbonyl B promoted 1,4-Michael-addition (Path b) is suggested to be less favored with a calculated Gibbs free energy of 18.7 kcal/mol for transition state cisoid-TS3 due to the ring constraint. Although 1,4-addition may produce the enolate oxygen-coordinated Mn(I)-intermediates cisoid-E (0.6 kcal/mol). Similarly, the introduction of 1a to cisoid-ECO initiates the proton transfer via transition state cisoid-TS5 (6.9 kcal/mol), leading to 3a after keto-enol tautomerism.

Fig. 8: Computed potential energy surfaces for the major reaction pathways of the Mn(I)-catalyzed carbon-skeleton rearrangement via C–C bond activation at the SMD(DCE)/M06L/Def2tzvp//M06L/def2svp level of theory.
figure 8

Purple line the DFT-computed energy surfaces of Path a. Blue line the DFT-computed energy surfaces of Path b.

Full size image

To further understand the driving force of the Mn(I)-catalyzed sigmatropic rearrangement, the localized Kohn-Sham orbitals and distortion/interaction analysis of the migratory insertion (C=C insertion) and the reverse β-aryl elimination (C=O insertion) were carried out. As shown in Fig. 9, the reverse β-aryl elimination reaction (C=O insertion) is driven by 1) the interaction between the C-Mn p-d orbital and the C=O π* antibonding orbital, and 2) the interaction between the C=O π orbital and the Mn dz2 antibonding orbital. In comparison, the C=C insertion is contributed by the donation of the C-M p-d orbital to the C=C π* orbital and the donation of the C=C π orbital to the antibonding dz2 orbital. In both cases, the energy gaps between the C-M p-d orbital and the π* orbital (3.10 eV for the C=O insertion, 3.05 eV for the C=C insertion) are smaller than those between the π orbital and the antibonding dz2 orbital (4.46 eV for the C=O insertion, 4.54 eV for the C=C insertion), suggesting the p-d to π* interaction attributes mainly for the insertion. Notably, the p-d to π* interaction is more significant in the C=C insertion (ΔEgap = 3.05 eV) than in the C=O insertion (ΔEgap = 3.10 eV), resulting in a better interaction energy (ΔEint = −14.3 kcal/mol) in the C=C insertion transition state transoid-TS2, which is probably the driving force of the preference of C=C insertion over C=O insertion for the rearrangement process. Furthermore, the orbital composition analysis in the LUMO of the substrates indicates that the terminal C4 p orbital (32.9%) in the C=C insertion step is greater than the carbonyl C2 p orbital (27.5%) in the C=O insertion step, resulting in a large driving force for the C=C insertion, guaranteeing the subsequent 1,3-STR. The difference in the interaction energies between transoid-TS2 (−14.3 kcal/mol) and cisoid-TS1 (−7.8 kcal/mol) thus resulted in a lower Gibbs free energy of activation for the C=C insertion transition state transoid-TS2.

Fig. 9: The localized Kohn-Sham orbitals and distortion/interaction analysis.
figure 9

a C-Mn p-d orbital and C=O π* antibonding orbital interaction, C=O π orbital and Mn dz2 orbital interaction. b Deformation/interaction analysis of the transition state for the β-aryl elimination pathway. c C-M p-d orbital and C=C π* orbital interaction, C=C π orbital and Mn dz2 orbital interaction. d Deformation/interaction analysis of the transition state for the migratory insertion pathway. Energies are in kcal/mol, isovalue of Kohn-Sham orbitalis is 0.05.

Full size image

Discussion

In conclusion, we developed a Mn(I)-catalyzed carbon-skeleton rearrangement via C–C σ bond cleavage. A variety of α-aryl-β, γ-unsaturated alcohols were applicable to this simple catalytic system. This protocol features with [1,n]-sigmatropic rearrangement (n = 2, 3), leading to 1,2-STR and 1,3-STR to reorganize carbon-skeletons with high atom- and step-economy. This sigmatropic rearrangement strategy can enrich the rearrangement chemistry and enhance the development of more challenging carbon-skeleton rearrangements in the future.

Methods

Procedure for the Mn(I)-catalyzed sigmatropic rearrangement of β, γ-unsaturated alcohols

All the reactions were run in an oven-dried 2 dram vial fitted with an oven-dried Teflon coated stir bar, and a Teflon cap under argon atmosphere. The Mn(CO)5Br that was used was stored in a freezer. Before every reaction set-up, DCE was freshly distilled. To a 10 mL vial equipped with a magnetic stir bar, was added α-aryl-β, γ-unsaturated alcohols (0.2 mmol), Mn(CO)5Br (5 mol %) and DCE (2.0 mL) under Ar atmosphere. The reaction mixture was then allowed to stir at 75 °C for 24 h. After the reaction mixture was cooled down, the corresponding reaction mixture was purified by flash chromatography on silica gel to afford the desired 1,2-STR and 1,3-STR products.