Introduction

Lactams are a class of N-heterocycle pharmacophores that figures prominently in medicinal chemistry. Representative skeletons, such as β-lactam1, pyrrolidine (γ-lactam)2,3, pyridinone and piperidinone (δ-lactam)4,5 etc., are all highly valued motifs underpinning many drugs and clinical entities (Fig. 1a). The 8-membered N-heterocycle, namely, azacine and its saturated form azocane, are present in important alkaloids6 and therapeutic agents7 but have not been extensively explored in medicinal efforts due to the paucity of general and modular synthetic strategies. Thus, the last decade has witnessed a rapid increase in its efficient preparation8.

Fig. 1: Background and research synopsis.
figure 1

a Representative lactam scaffolds in drugs and bioactive natural products. b Reaction mode of traditional ketene chemistry. c Ring-opening of electron-deficient cyclopropene in synthetic application. d Functionalized alkenyl ketene via group-migrating ring-opening of “D-A”-type cyclopropene reveals vinylogous cyclization mode (this work).

Full size image

Since its first documentation by Staudinger in 1905, the ketene chemistry has evolved into an indispensable synthetic arsenal for the access to various N- and O-heterocycles and functionalized carboxylic acid derivatives9,10,11,12. Ketene reacts either directly with nucleophiles (type I), or, under the umpolung catalysis of a (chiral) Lewis base (type II), turns into an enolate-like species to engage electrophiles into cyclization products, which has evolved into a powerful strategy for chiral heterocycle synthesis. Despite of the significant progresses, the commonly used ketene is not sufficiently electrophilic to engage less reactive nucleophiles in type I reactivity, such as sterically bulky and/or less nucleophilic ones. Consequently, to our knowledge, the Staudinger [2 + 2] reaction of ketene with imine, a highly useful strategy for β-lactam scaffold synthesis, has not yet been applicable to ketoimines. As a consequence, except in a few specialized cases13,14, the tetra-substituted β-lactam chemical space remains largely untapped due to the limitations of currently available β-lactam synthetic strategies15. More importantly, although methods for ketene generation have been continuously developed, such as the well-established Wolff rearrangement of diazo compounds16, photoredox deoxygenated ketil17, carbonylation of carbene or metal carbenoid18 and ylide anion19, as well as oxidation of alkylidene carbenoid20, etc., the reactivity of ketenes derived from all known methods are still limited to the conventional [2+n] mode where ketene is used as a 2 C synthon (Fig. 1b). Innovative methods offering structurally distinct ketene intermediates would be highly needed to access currently intangible chemical space in heterocycle synthesis via a different paradigm. For instance, the vinylogous coupling cyclization of alkenyl ketene as a 4 C synthon would be an alternative reactivity mode aside of the conventional [2 C + n] mode, which remains elusive due to unfavored periselectivity21,22.

Cyclopropene (CPE) has garnered intensive recent attention from the synthetic and theoretical communities due to its ready availability, high thermostability and versatile reactivity profile23,24,25. Electron-deficient cyclopropenes can be easily accessed via metal-catalyzed or photochemical cyclopropanation of alkyne, but in contrast to the “donor-acceptor cyclopropane” (DAC) chemistry26,27,28,29,30, which gained great momentum over the past two decades for carbo- and heterocycle synthesis, the ring-opening of electron-deficient cyclopropene as “dehydro”-type DAC (DDAC) substrate has special complications. Owing to its higher strain (54 kcal/mol) as compared to cyclopropane (27 kcal/mol), they have lower-lying back-bone σ*sp2-sp3 orbitals (LUMO), resulting in very facile ring-opening via SN2 type Lewis base attack at the less-sterically hindered backbone Csp2‒Csp3 σ bond with or without a Lewis acid (Fig. 1c, path a)31,32,33. In contrast, ring-opening at the more substituted alkene carbon site involving an intermediate with an sp2-hybridized alkene cationic center, referred to here as the SN1-type ring opening (path b), remains highly under-developed. In fact, the only example of this type was reported by France et al. in 2012, wherein the putatively formed highly unstable sp2-carbocation-containing zwitterionic intermediate I was intramolecularly intercepted by a pendant electron-rich heterocycle to afford bicyclic phenol derivatives (Fig. 1c, path b, eq. 1)34. Not surprisingly, the donor moiety (R3 in I) can only be groups that can significantly stabilize the carbocation, such as electron-rich aryl and heterocycle, R3SiCH2-, etc. Based on our long-standing interest in cyclopropene chemistry35,36,37,38,39,40, herein we report a unique Lewis acid catalyzed SN1-type ring-opening of readily available 3,3-diester substituted cyclopropene, which generates a functionalized alkenyl ketene intermediate III via 1,4-alkoxy migration of the incipient sp2-carbocation-containing zwitterionic intermediate II (Fig. 1c, path b, eq. 2). This highly electrophilic functionalized ketene intermediate not only enables challenging nucleophiles to be engaged in the conventional [2+n] reactions, granting access to hithertofore challenging tetra-substituted β-lactam chemical space in Staudinger [2 + 2] reaction, but also opened the vinylogous [4+n] cyclization mode, as exemplified by the [4 + 1] and [4 + 4] cyclizations to pyrrolidinone and azocine frameworks, respectively (Fig. 1d). Our protocol displays a wide donor scope (vide infra) and significantly expands the synthetic application of the DDAC system for further development.

Results

[2 C + 1 N] reaction with iminophosphoranes

Our study commenced with the [2 C + 1 N] reaction of cyclopropene 1a with N-Ph iminophosphorane 2a. After a brief optimization we found that the reaction can deliver ketene imine 3a in a high yield of 80% when reacted with phenyl iminophosphorane at 50 °C under the catalysis of 20 mol% Yb(OTf)3 in dichloromethane (DCM). The conditions turned out to be general for an array of representative iminophosphoranes (Fig. 2). Variation of the donor group on the cyclopropene shows that electron-deficient aryls are well-tolerated. Even 4-CF3 substituted substrate was competent, despite of the low 27% yield of product 3n likely arising from slower formation and higher reactivity of the ketene. Notably, the substrate bearing an electron-rich 4-MeC6H4 substituent is also competent, albeit with lower efficiency (3o). Surprisingly, 4-MeOC6H4 substituted cyclopropene resisted to undergo ring-opening even at elevated temperatures (3p), with most cyclopropene being recovered. This stands in stark contrast to the canonical “donor-acceptor” system as reported by France et. al. 34, where only electron-rich donors are allowed. Nonetheless, both linear (3s) and branched (3t) alkyl substituents were found feasible, although the latter has a lower yield likely arising from the lower reactivity of the ketene intermediate. The structure of the ketene imine products is confirmed by the X-ray of dibromo analogue 3u. The migrating moiety could also be ethoxy (3v), neopentoxy (3w) and phenoxy (3x), with only slightly diminished efficiencies. The lower yield of 3x is also likely the result of unfavorable electronic perturbation on the reactivity of the ketene. Unfortunately, substitution at both olefinic carbons of the cyclopropene is not compatible (3y-3z).

Fig. 2: The scope of [2 + 1] reaction with iminophsphoranes.
figure 2

a Scope of iminophosphorane. b Scope of cyclopropene.

Full size image

[2 C + CN] reaction with imines and [2 C + CNN] reaction with azomethine imines

The Staudinger [2 C + CN] reaction with imines 4 turned out to be highly yielding for N-aryl aldimines under catalytic action of Fe(OTf)2, regardless of electronics (5a-5c), heteroatoms (5e-5g), alkenyl (5 h) or N-substituents (5i-5j), consistently affording the trans-isomers exclusively or predominantly (Fig. 3a). Encouraged by the high reactivity, we attempted ketoimines, which are less reactive due to steric hindrance in the cyclization step. Remarkably, under slightly modified conditions, the reactions of ketoimines delivered the expected tetra-substituted β-lactams in moderate to good yields (Fig. 3b), regardless of the electronic properties of the imine aryl moiety (5k-5r), and accommodating heterocyclic imines (5s-5v), and alkyl beyond methyl (5w). Note that the electron-rich heterocycle derived imines afforded 5s-5u with cis-isomers as major products, presumably due to post-reaction epimerization as caused by these stronger heterocycle donors, which makes in-situ “donor-acceptor” type ring-opening/re-annulation more feasible. Spiro-lactams have attracted much attention due to their great potential in drug development41. N-phenyl-9-fluorenyl imine can be facilely converted to the spiro-lactam 5x in a high yield of 87%. X-ray diffraction studies of cis-5a and trans-5k revealed their extraordinarily long Cα‒Cβ bond of 1.59 and 1.60 Å, respectively, attributed to the “donor-acceptor type” substitution pattern (Fig. 3). Moreover, the [2 C + CNN] type reaction with azomethine imine 6, previously only realized in a polarity-reversed setting acting as an electrophile due to its low nucleophilicity42, can also be realized in a highly yielding manner as demonstrated by the several representative examples (Figs. 3c, 7a-e).

Fig. 3: Scope of the [2 + 2] and [2 + 3] cyclization.
figure 3

a The Staudinger [2 + 2] reaction with aldimines. b The Staudinger [2 + 2] reaction with ketoimines. c The [2 + 3] cyclization with azomethine imines.

Full size image

Formal periselectivity issue in the reaction with α,β-unsaturated imines

Instead of the expected [2 + 2] product, under the catalysis of Fe(OTf)2, (E)-α,β-unsaturated ketoimine 8 bearing electron-neutral to electron-deficient aryl moieties reacted with cyclopropene 1a to afford the 8-membered lactams (9a-9f) in good yields and high diastereoselectivites via a [4 + 4] annulation (Fig. 4). Notably, for the 4-MeO-2,3,5,6-tetrafluoro-substituted substrate, azocine 9 g was formed along with trans-divinyl-β-lactam 10 g (X-ray). In contrast, for the electron donating 4-MeO-substituted α,β-unsaturated imine, azocine 9 h was obtained along with a small amount of the [2 + 4] product 11 h. In a separate reaction, after full conversion of the imine the reaction mixture was subjected to 80°C, 9 h was completely converted to 11 h in 79% yield, suggesting that 8-membered lactam is a kinetic product as compared to the δ-lactam. 2-Thienyl substituted α,β-unsaturated imine indeed furnished the six-membered lactam 11i exclusively owing to the electron-donating nature of thiophene, so is the 4,4-disubstituted α,β-unsaturated imine, as shown in 11j. The structure and configuration of the 8-membered and 6-membered lactams were confirmed by the X-ray diffraction studies of 9b and 11j, both also featuring a relatively long Csp3‒Csp3 bond (1.56 and 1.58 Å respectively) at the ring closure site. When α,β,γ,δ-unsaturated ketone imine was used as the coupling partner, the corresponding azocine 9k and β-lactam 10k were formed in 25% and 38% yield, respectively. The 4-(2,6-dimethyl)phenyl substituted α,β-unsaturated imine only afforded the [2 + 2] product 10 l in a high yield and a low diastereomeric ratio (dr). Isolated 10 l, when treated with Yb(OTf)3 at 60 °C, sluggishly converted to the corresponding 6-membered isomer 11 l in a low yield with concomitant O-demethylation, while most remained 10 l became O-demethylated to the corresponding ketone. The above results suggest that the [2 + 2] cyclization precedes the formation of azocines 9, with the latter likely arising from ring-enlarging Cope-rearrangement of the initially formed β-lactam via “donor-acceptor” type ring-opening43. The substituent R2 of the α,β-unsaturated imine could also be alkyl, such as t-Bu (9 m) or cyclopropyl (9n/10n/11n). Upon variation of aryl moiety of the cyclopropene, we obtained 9o-9q in good yields along with the corresponding tetra-substituted dialkenyl β-lactams 10o-10q as minor isomers. Finally, 1-n-butyl cyclopropene was also converted to the corresponding azocine and β-lactam products (9r/10r) albeit with a lower efficiency due to competing side processes in the cyclization stage. If desired, these β-lactam products could be converted to the 8-membered or 6-membered products under controlled conditions facilitated by Lewis acid, but care should be taken in each case to prevent over-conversion to the six-membered lactam if the azocines are desired products (such as 11o).

Fig. 4: Scope of the reaction with α,β-unsaturated imines.
figure 4

All products were obtained with > 20/1 dr unless specified otherwise.

Full size image

Exclusive [4 C + 1 N] periselectivity in the reaction with 2H-azirines

Highly oxygenated γ-lactams are featured in a class of terpenoid natural products which display a diverse array of important biological activities due to their multiple electrophilic sites3 (Fig. 5a). The epoxy-γ-lactam epolactaene, a microbial metabolite and its analogues are exemplifying, which induces apoptosis in human leukemia B-cell line, BALL-1. Structure activity relationship (SAR) study of synthetic epolactaene derivatives by Kobayashi et. al. suggests that the carbonyl epoxy-γ-lactam scaffold is key to their activities44. We found that upon treatment with various 2H-azirines 12 bearing various (hetero)aryl and alkyl substituents, the ketene generated from 1a under the catalysis of Yb(OTf)3 facilely retrieved exclusively the [4 C + 1 N] products 13a-f, 13i, which bear the unsaturated γ-oxygenated γ-lactam scaffold, in good yields albeit low Z/E ratios (Fig. 5b). Their structure was confirmed by X-ray crystallographic studies of Z-13i. The γ-alkyl analogue 13h could also be obtained with a lower efficiency due to competing side processes but surprisingly, a high Z selectivity. Upon treatment with H2O2 and NaOH, product (Z)-13a was facilely converted to the key epoxy-acid 14 that contains the bicyclic epoxy-γ-oxy-γ-lactam framework, in a high yield and 3/1 dr, showing the high usefulness of the [4 + 1] protocol.

Fig. 5: The natural occurrence of γ-oxy-γ-lactam scaffold and its facile construction via the vinylogous [4 + 1] cyclization.
figure 5

a Representative γ-oxy-γ-lactam type natural products; b The scope of the [4 + 1] cyclization with azirines and product epoxidation.

Full size image

We performed preliminary in-vitro antiproliferative evaluation of selected β-, γ-, δ-lactams and azocines obtained above at 40 μM concentration against 5 strains of human cancer cell lines (HL60, MDA-MB-231, hepG2, A549 and K562), and the results show that all scaffolds display appreciable activities to various degrees, among which δ-lactam 11j gave the highest 61.1 ± 11.4% (Supplementary Table 2), and tetra-substituted β-lactams display higher growth inhibition over the tri-substituted ones in part of the cell lines (K562 and A549) and comparable activities in the rest strains, highlighting the advantages of tetra-substituted β-lactams in improving efficacy. These results highlight the therapeutic potential of these lactams.

Synthetic maneuverer of lactam products

The synthetic utility of this transformation is further illustrated in a few examples below (Fig. 6). First, treatment of β-lactam 5a with Zn/AcCl efficiently removed O-methyl group and converted it to the ketone 15 (Fig. 6a). When the γ-lactam product 13i (Z/E = 1/1) was treated with TsOH•H2O at 80 °C, a hydrative rearrangement occurred to deliver product 16a in a moderate yield (Fig. 6b). The product of a substrate with unsymmetrically substituted enamine moiety indicates that the aryl proximal to the nitrogen undergoes migration, based on which a possible mechanism involving 1,2-N migration is provided in Supplementary Fig. 2. The 8-membered lactam 9a can be synthesized on a gram scale in a good yield of 70% (Fig. 6c, i). Saponification of the ester and coupling with N,O-bismethyl hydroxyamine afforded the corresponding Winreb amide 17 in 45% overall yield (Fig. 6c, ii). Treatment of 7a with LDA in THF also retrieved its N,N-diisopropyl amide 18 (Fig. 6c, iii). Intriguingly, when treated with Zn/AcCl (4 equiv.), 9a was converted through a multi-stage process to a highly complex tetracyclic fused β-lactam 19 in 57% yield (Fig. 6c, iv), whose structure was also determined by single crystal X-ray diffraction, by way of slow methanol elimination of its precursor, intermediate 20 (Supplementary Fig. 7). The latter was rapidly assembled presumably via a SET reduction/radical cyclization/oxidation/radical cyclization/Hydrogen atom abstraction (HAA) sequence (a detailed description of the mechanism is shown in Supplementary Fig. 8). These results highlight the significance of the methoxy group in fostering diversified reactivities.

Fig. 6: Product elaborations.
figure 6

a Methyl ether deprotection of β-lactam. b Hydroytic rearrangement of γ-oxy-γ-lactam. c Gram-scale synthesis of azocine 9a and its downstream conversions.

Full size image

Experimental and theoretical mechanistic studies

Some mechanistic studies were then performed to elucidate the mechanism of the unusual ring-opening (Fig. 7). The reaction of 1a with nitrone 21 in a one-pot reaction did not yield the [2 + 3] product; rather, the [2 + 2] product 22 was obtained in a good yield with the oxygenated side product 23 (Fig. 7a). The latter arises from trapping of the initially formed zwitterionic intermediate from the donor-acceptor ring-opening by nitrone followed by fragmentation. The in-situ released N-methyl imine then participated the [2 + 2] reaction. In contrast, preforming the ketene intermediate III and adding nitrone afterwards led to formation of the 5-membered 1,2-oxazolone 24 as a single diastereomer (Fig. 7a). Treatment of 1a with catalytic Fe(OTf)2 in trifluoroethanol indeed resulted in the trapping of the initially formed zwitterionic intermediate II (product 25) and the methoxy-migrated ketene intermediate III (product 26) (Fig. 7b). A crossover experiment between 1b and 1c did not result in alkoxy scrambling, and the products 5 y and 5z were obtained as the sole products (Fig. 7c), suggesting that under the catalytic conditions the alkoxy migration occurred in an intramolecular fashion. Examination of the Staudinger [2 + 2] reactions with aldimine in a few solvents shows that coordinative tetrahydrofuran (THF) inhibited catalysis, weakly coordinating 1,2-dichloroethane (DCE) and toluene are only marginally viable, while DCM turns out the most suitable (Supplementary Table 3). The reaction of cyclopropene 1r, with one ester being replaced by a sulfone group, resulted in exclusive formation of the formal 1,2-sulfonyl migration product 1 u (Fig. 7d)45. Other 3,3-disubstitution, like dicyano- and 3-CF3-3-CO2Me-substituted cyclopropene, also failed to undergo ring-opening (Fig. 7e), suggesting the importance of chelation for effective activation.

Fig. 7: Experimental mechanistic studies.
figure 7

a Distinct outcome of one-pot and sequential addition methods. b Trapping of the primary and secondary intermediates with trifluoroethanol. c Cross-over experiments. d The reactivities of other electron-deficient cyclopropenes.

Full size image

To gain further insight into the ring-opening step of this “dehydro” DAC system and rationalize the unusual electronic requirement on the donor moiety, we performed DFT calculations with cyclopropene 1a and 1 h and Sc(OTf)3 as the catalyst due to its high catalytic activity in the formation of ketene and lower computational cost (Fig. 8a). Evaluating the coordination modes suggested that chelation of substrate by the two carbonyl oxygens (O1 and O2) is most stable. In initial gas-phase optimization without solvation model the ring-opening transition structure (TS) and formed zwitterionic intermediate could not be located, while incorporation of an explicit solvent DCM molecule enabled location of these unstable structures. To better describe the solvation effect, further calculation adopted a hybrid solvation model with the implicit SMD model46 and the explicit DCM molecule in view of the uniqueness of this solvent observed experimentally (Supplementary Table 3). From the carbonyl chelated 1a•Sc(OTf)3 adduct Com1, SN1 type ring-opening has a small activation barrier of 12.8 kcal/mol, and can lead to two products due to the presence of the DCM molecule, depending on the orientation of the latter. In path 1, where the Cl atom in DCM served to stabilize the cationic center, ring-opening of the cyclopropene framework leads directly to five-membered ring closure (to Int1) and subsequent O-methyl epimerization to the highly conjugated cyclic intermediate Int2 (Fig. 8a, path 1). If the Cl in DCM is not oriented to stabilize the cationic center, the short-lived cationic intermediate Int1a may result (intermediate II in Fig. 1). Then it may undergo a barrierless cyclization via TS2a to the relatively less stable Int2a. Rationalization of the proximity of DCM to the cationic center can be gleaned from high level ab initio molecular dynamics (AIMD) simulation with Orca 5.047,48 (Supplementary Video 1; representative snapshots are provided in Supplementary Fig. 29). These results highlight the transient nature of the initially formed zwitterionic intermediate, and the effect of solvent molecule on its fate. As corroborated experimentally, the intermediate can only be intercepted with stoichiometric negatively charged nitrone oxygen, or excessive, solvent level trifluoroethanol, while in DCM, intramolecular alkoxy migration resulted (Fig. 7, a ~ c). For the SN1 type ring opening of cyclopropene 1 h bearing a strong donor 4-MeOC6H4 moiety, ring-opening is 2.4 kcal/mol more favored than 1a (TS1’ vs TS1), leading to a more stabilized sp2-carbocation-containing intermediate Int1’ (by 2.9 kcal/mol compared to Int1). Int1’ then undergoes fast ring-closure via TS2’ (4.6 kcal/mol barrier) to afford the cyclic intermediate Int2’ which has very similar in energy with Int1’ (Fig. 8a, path 2). Till this stage, for substrate 1 h, the initial ring-opening is more kinetically favored and thermodynamically accessible. It is thus likely the subsequent steps or other factors that makes this substrate incompetent.

Fig. 8: DFT studies on the formation of ketene.
figure 8

a Gibbs free energy profile (energies in kcal/mol) and geometries (bond lengths in Å) of the group-migrating SN1 type ring-opening of cyclopropene. b Noncovalent interaction analysis of the cyclic intermediates (blue, strong attraction; green, weak interaction).

Full size image

The bond lengths of C4‒O3 and C1‒O3 in Int2 are both 1.48 Å, slightly shorter and longer respectively than those of C4‒O3 (1.49/1.51 Å) and C1‒O3 (1.42/1.43 Å) in Int2a/Int2´, indicative of a better stabilization of the cationic center in Int2. The multiple interactions (triple C‒H•••O, C‒ H•••Cl and Cl•••π) between solvent and Sc(OTf)3 as well as the O-Me group of substrate in Int2 as compared to the fewer interactions (double C‒H•••O) in Int2´may contribute to the more stabilization of the former. This is further corroborated by non-covalent interactions (NCI) analysis: the larger green-colored regions and DCM stabilizes the carbocation (C1) of the five-membered ring via Cl•••π interactions in Int2 than in Int2´ and Int2a (the red cycle in Int2, Fig. 8b). NBO charge analysis also suggests that the hydrogen-bonded CH2Cl2 serve to stabilize the cationic center in Int2 to a larger extent as compared to Int2´ and Int2a by the chlorine atom through the aid of OTf of the Lewis acid (Supplementary Fig. 28). Therefore, the calculation results highlight the significance of solvent participation in the ring-opening process.

Following the cyclization, elongation of the Sc‒O1 bond (red) in Int2 triggers MeO—Csp bond scission, completing the methoxy group migration event. The formed Lewis acid-ketene adduct Int3 is found to be > 10 kcal/mol less stable than Int2, suggesting that if other factors are not involved to complicate the scenario (such as involvement of extra Lewis acid), the zwitterionic form should be more dominant form. Inspection of the crude mixture of the ring-opening reaction mixture, however, did not reveal conclusive evidences (Supplementary Fig. 13), which awaits further studies.

In this work, we present the generation of a functionalized alkenyl ketene intermediate from a unique Lewis-acid-catalyzed alkoxy migrating SN1-type ring-opening of the readily available electron-deficient cyclopropenes. Upon coupling with a series of N-nucleophiles, the electrophilic ketene intermediate demonstrates vinylogous [4+n] cyclization mode, as exemplified by the [4 + 1] and (formal) [4 + 4] cyclization toward oxygenated γ-lactam and azocines, besides the conventional [2+n] reactivities. The suite of lactam products display significance as anti-cancer candidates, natural product scaffold or precursors to other interesting N-heterocycle scaffolds. Mechanistic studies suggest that the alkoxy migration occurs in an intramolecular fashion from the initially formed transient sp2-carbon-containing zwitterionic intermediate. DFT provided hints on the facileness of the ring-opening, revealing the broad synthetic potential of the under-developed “dehydro” donor-acceptor cyclopropane (DDAC) system. Further studies on this system might be an important direction of the donor-acceptor system chemistry.

Methods

General procedure for the Yb(OTf)3-catalyzed [2 + 1] reaction of cyclopropene with iminiophosphorane 2 and [4 + 1] reaction with 2H-azirine 8 (one-pot protocol):

To a dry Schlenk pressure tube was added powered 4 Å molecular sieves (200 mg), and the tube torched with a heat gun (gun temperature: 600 °C) under vacuum to remove the moisture. The tube was cooled to room temperature, Yb(OTf)3 (0.04 mmol, 20 mol%), cyclopropene 1 (0.24 mmol), and iminophosphorane 2 or 2H-azirine 12 (0.20 mmol) and DCM (2 mL) were added in one pot, the tube was sealed, and the mixture was stirred at 50°C (for iminophosphorane) or room temperature (for 2H-azirine) until completion (TLC). Then the solvent was removed in vacuo and the residue submitted to flash chromatography to obtain the products.

For the reaction of 1-alkyl cyclopropene with iminiophosphorane 2, a step-wise protocol was adopted: to a dry Schlenk pressure tube was added powered 4 Å molecular sieves (200 mg) and Yb(OTf)3 (0.04 mmol, 20 mol%), and the tube torched with a heat gun (gun temperature: 600 °C) under vacuum to activate the catalyst. The tube was cooled to room temperature, cyclopropene 1 (0.24 mmol) and DCM (2 mL) were first added, the tube was sealed, and the mixture stirred at 50 °C until full conversion to ketene. Then, iminophosphorane 2 (0.20 mmol) were added and the mixture was heated at 50 °C under Ar until complete conversion of the latter (TLC). Following the same workup as above to obtain the products.

General procedure for the Fe(OTf)2-catalyzed coupling cyclization of cyclopropene with imines 4 and α,β-unsaturated ketoimine 8:

A dry Schlenk pressure tube was charged with Fe(OTf)2 (0.06 mmol, 30 mol%) and 4 Å molecular sieves (200 mg), the tube was torched with a heat gun (gun temperature: 600 °C) in vacuo to activate the catalyst. After cooling to room temperature, for the one-pot protocol (for aldimines and most of the ketoimines), cyclopropene 1 (0.30 mmol), imine 4 (0.20 mmol) and DCM (2 mL) were added. The tube was sealed and mixture was heated at 60 °C under Ar until complete conversion (typically 2 days). For step-wise protocol (for ketoimines 4r, 4 v and α,β-unsaturated ketoimines 8), 1 (0.30 mmol) and DCM (2 mL) were first added, and the mixture stirred at to 60 °C under Ar until complete conversion to ketene (TLC). Then imine 4r, 4 v or 8 (0.20 mmol) was added and the mixture was stirred at the substrate-specific temperature and conditions until complete conversion of the imine. The conditions specified in the Supplementary Information is advised to be followed for each case. Finally, the solvent was removed in vacuo and the residue purified by flash chromatography to obtain the corresponding products.

General procedure for the In(OTf)3 catalyzed [2 + 3] reaction of cyclopropene with azomethine imines 6 (step-wise protocol):

To a dry Schlenk pressure tube was added In(OTf)3 (0.06 mmol, 30 mol%) and 4 Å molecular sieves (200 mg). The tube was torched with a heat gun (gun temperature: 600 °C) in vacuo to activate the catalyst. After cooling to room temperature, cyclopropene 1 (0.40 mmol, 2.0 equiv.) and DCM (2 mL) were added. The tube was sealed and the mixture was heated at 60 °C under Ar until complete conversion to ketene (TLC). Then azomethine imine 6 (0.20 mmol) was added to the solution and the mixture heated at 50 °C until complete reaction (TLC). The solvent was removed in vacuo and the residue submitted to flash chromatography (DCM:EA = 50:1 to DCM:EA = 20:1) to obtain the products 7.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.