Difunctionalization of bicyclo[1.1.0]butanes enabled by merging C−C cleavage and ruthenium-catalysed remote C−H activation

Main

In the quest of ‘escaping from flatland’^{1,2,3,4,5,6,7,8}, highly functionalized cyclobutanes with unique puckered linear geometry have received increasing attention in medicinal chemistry. The fraction of sp³-hybridized carbon atoms (Fsp³) was identified as a key descriptor of drug likeness, and the increased Fsp³ in cyclobutanes renders them as a privileged motif for isosteres in drug design⁹. The incorporation of cyclobutane scaffolds often enhances the physicochemical and pharmacokinetic properties of drug molecules. In contrast to usually flat arene rings, polysubstituted cyclobutanes with high Fsp³ can provide, among others, improved solubility due to their nonplanar substituent vectors. Indeed, several pharmaceuticals featuring 1,3-bifunctional cyclobutanes have been clinically tested, such as PF-03654746 (ref. ¹⁰), NVP-ADW 742 (ref. ¹¹), NK1 selective antagonists¹² and Linsitinib¹³, translating into a strong need for innovative cyclobutane syntheses (Fig. 1a). However, in contrast to rather well-established bifunctionalization of bicyclo[1.1.0]pentanes (BCPs)^{14,15,16,17,18} and ring expansions of bicyclo[1.1.0]butanes (BCBs)^{19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41} (Fig. 1b), strategies to access structurally complex cyclobutanes are, unfortunately, scarce. Thus far, these syntheses are limited to radical or nucleophilic additions to BCBs, typically resulting in mono- or disubstituted cyclobutanes^{42,43,44,45,46,47,48,49}, with only few examples for densely substituted cyclobutanes^{50,51,52,53,54}. Furthermore, the aforementioned transformations are largely limited to rather harsh conditions and elements of prefunctionalization on the coupling partner, while the merger of C–C activations^{55,56,57,58,59,60} of BCBs with remote C–H activation⁶¹ has thus far proven elusive. To address these current topical needs, we have now developed the merger of BCBs C–C functionalization with remote C–H functionalization^{62,63,64,65,66,67,68,69,70,71,72,73} by a single, yet powerful, ruthenium(II) catalyst. Salient features of our findings include (1) versatile 1,3-difunctionalization of BCBs for tri- and tetrasubstituted Fsp³-rich cyclobutanes; (2) a single ruthenium complex for a multi-catalysis manifold including an efficient ruthenacycle-mediated halogen-atom transfer (Ru-XAT) process, C–C scission and meta-C–H functionalization; (3) exceedingly mild reaction conditions and (4) outstanding chemo- and site selectivities (Fig. 1c).

**Fig. 1: Design blueprint for the difunctionalization of BCBs to access 1,1,3-trisubstituted and 1,1,3,3-tetrasubstituted cyclobutanes enabled by remote C–H activation.**

Results and discussion

Optimization studies

We initiated our studies on BCBs 1,3-difunctionalization through remote C–H activation, with benzyl-bicyclo[1.1.0]butane-1-carboxylate (1a), 2-phenylpyridine (2a) and ethyl-2-bromo-2,2-difluoroacetate (3a) as the model substrates (Table 1). We were pleased to find that the desired product 4 was efficiently obtained in 75% isolated yield with Ru(O₂CMes)₂(p-cymene) as the catalyst and P(4-CF₃C₆H₄)₃ (Table 1, entry 1). Next, a series of phosphines, as well as alternative ligands such as bipyridine (bpy) and N-heterocyclic carbene (NHC), were tested, and P(4-CF₃C₆H₄)₃ was found to be superior (Table 1, entries 2 and 3). Ru(OAc)₂(p-cymene) gave the desired product 4 with 52% yield, while [RuCl₂(p-cymene)]₂ was ineffective (Table 1, entry 4), highlighting the importance of carboxylate assistance in the C–H ruthenation⁷⁴. [Ru(^tBuCN)₅(H₂O)](BF₄)₂⁷⁵ as a precatalyst gave inferior results (Table 1, entry 5). Control experiments revealed the crucial role of the ruthenium catalyst and the phosphine ligand for the BCBs C–C cleavage difunctionalization (Table 1, entry 6).

Table 1 Optimization of the reaction parameters

Full size table

With the optimized reaction conditions in hand, we subsequently evaluated the viable substrate scope for the bifunctionalization of BCBs 1 with differently substituted heteroarenes 2 (Fig. 2). Arenes with distinct electronic features and substituents, such as fluorine, thioether, bromine, ester and keto groups, were fully tolerated by the versatile catalyst (5–29). Transformable pyrazoles, ketimine and oxazolines could be employed to guarantee meta-selectivity (17–24, 30). Drug-relevant motifs, such as diazepam, purines and nucleoside proved to be viable for the ruthenium-catalysed BCBs difunctionalization (26–29).

A series of substituted BCB esters 1 furnished the 1,1,3-trisubstituted cyclobutanes 31–37 (Fig. 3). Thus, BCBs featuring sensitive functional groups, including ester, thiophene, ketone, amide and sulfone (38–41) were efficiently converted to desired products. Furthermore, disubstituted BCBs were identified as amenable substrates (42–47). Here, disubstituted BCBs favoured to form a benzylic radical rather than a tertiary radical, chemo-selectively delivering diarylcyclobutane motifs. Likewise, a wide range of alkyl halides, such as perfluoroalkane halides, fluorine-free alkyl bromide, monofluoroalkyl bromide and difluoroalkyl amides were tolerated to give the desired cyclobutanes 48–57.

**Fig. 3: Scope of BCBs and radical precursors.**

Gram-scale and late-stage derivatization

To demonstrate the practical utility of our BCBs C–C scission/remote activation strategy, cyclobutane 4 was prepared at gram scale with comparable efficacy (Fig. 4a). The site selectivity ensured that pyrimidine, oxazoline and ketimine (vide infra) as well as pyridine and pyrazole were efficiently diversified, extending the viable portfolio (Fig. 4b). Further, a triple activation manifold proved viable in terms of ruthenium(II)-catalysed C−C/meta-C−H and ortho-photo-induced C−H activation (58 and 59).

DFT calculation and mechanism studies

To gain insights into the reaction mechanism, the site selectivity of radical addition at the two possible BCB sites was probed by means of density functional theory (DFT) calculations (Fig. 5a; see Supplementary Figs. 1, 2, 7 and 8 for details). In the case of monosubstituted BCB 1a, the difluoroalkyl radical preferentially attacked at the unsubstituted carbon, leading to the formation of a more thermodynamically stable tertiary radical. The shorter bridge C−C distance (1.61 Å) and the longer C−C_RF distance (2.21 Å) in TS3-s4, compared with TS4-s2, indicated that the radical attack at the unsubstituted site proceeded through an earlier transition state that structurally resembled the starting BCB 1a. In the case of radical addition to the disubstituted BCB 1m, the ester-substituted site is favoured, resulting in the formation of thermodynamically stable benzylic radical. During the cleavage of the σ-bridge bond in disubstituted BCB 1m, TS7-s1 exhibited an earlier transition state characteristic, evidenced by the slightly elongated C–C bridge distance of BCB 1a (1.63 Å in TS7-s1 and 1.64 Å in TS8-s1) and the relatively larger C–C distances between the BCB 1m and the difluoroalkyl radical (2.25 Å in TS7-s1 and 2.21 Å in TS8-s1). The presynthesized p-cymene-free ruthenacycle 66 yielded the desired product 4 in the presence of MesCO₂H and phosphine ligand (Fig. 5b). The control experiment without a phosphine ligand as well as with alternative ligands, such as bpy or NHC, in the optimization table failed to provide the desired product, highlighting the essential role of the phosphine ligand assistance^76,77,78,79. The key role of the phosphine ligand was further demonstrated by attempting this transformation with carboxylate-free ruthenium(II) phosphine complex 67, yielding the desired product in high yield when MesCO₂H was added (Fig. 5c). The carboxylate-free ruthenium(II) phosphine complex 67 could afford the desired product only in 14% yield even in the absence of MesCO₂H (Fig. 5c). Additionally, radical trapping experiments validated the radical mechanism (see Supplementary Data 7.1.1 for details). On the basis of our mechanism and computational findings, a plausible catalytic cycle is put forwards in Fig. 5d, featuring ruthenacycle A in a ruthenacycle-mediated XAT process to furnish ruthenium(III) intermediate B with an energy barrier of 14.9 kcal mol⁻¹. The thus formed alkyl radical 3^R attacks the BCB 1a at the unsubstituted site to induce strain-release C–C scission with an energy barrier of 15.9 kcal mol⁻¹, forming the tertiary radical C. Species C reacts at the ruthena(III)cycle B para to the C_Ar–Ru bond, delivering the stabilized singlet metallacycle D. Here, the formation of the cis product is favoured over the trans product by 1.4 kcal mol⁻¹. The subsequent rearomatization and proto-demetallation releases the desired cyclobutane 4, thereby regenerating the catalytically active ruthenium(II) species F.

Conclusions

We have achieved the merger of BCBs C–C activation with remote meta-C–H functionalization by a multipotent ruthenium catalyst. The double activation proceeded in a highly chemo- and position-selective fashion and provided access to densely decorated Fsp³-rich cyclobutanes in a single step. Mechanistic studies were suggestive of a Ru-XAT process enabling chemo-selective BCBs opening and meta-diversification.

Methods

General methods for 1,3-difunctionalization of BCBs enabled by ruthenium-catalysed remote C–H activation

The general procedure for 1,3-difunctionalization of BCBs was as follows: [Ru(O₂CMes)₂(p-cymene)] (16.8 mg, 10.0 mol%), P(4-CF₃C₆H₄)₃ (14.0 mg, 10.0 mol%), Na₂CO₃ (64 mg, 0.6 mmol, 2.0 equiv.), 1,4-dioxane (2.0 ml), 1 (0.3 mmol, 1.0 equiv.), 2 (0.9 mmol, 3.0 equiv.) and 3 (0.9 mmol, 3.0 equiv.) were added into an oven-dried 20 ml pressure tube. The reaction mixture was stirred at 65 °C for 24–48 h. After cooling to ambient temperature, the mixture was purified by column chromatography on silica gel to afford the corresponding cyclobutanes 4–57.

Whether it is listening to a piece of music, learning a new language, or solving a mathematical equation, people often acquire abstract notions in the sense of motifs and variables—manifested in musical themes, grammatical categories, or mathematical symbols. How do we create abstract representations of sequences? Are these abstract representations useful for memory recall? In addition to learning transition probabilities, chunking, and tracking ordinal positions, we propose that humans also use abstractions to arrive at efficient representations of sequences. We propose and study two abstraction categories: projectional motifs and variable motifs. Projectional motifs find a common theme underlying distinct sequence instances. Variable motifs contain symbols representing sequence entities that can change. In two sequence recall experiments, we train participants to remember sequences with projectional and variable motifs, respectively, and examine whether motif training benefits the recall of novel sequences sharing the same motif. Our result suggests that training projectional and variables motifs improve transfer recall accuracy, relative to control groups. We show that a model that chunks sequences in an abstract motif space may learn and transfer more efficiently, compared to models that learn chunks or associations on a superficial level. Our study suggests that humans construct efficient sequential memory representations according to the two types of abstraction we propose, and creating these abstractions benefits learning and out-of-distribution generalization. Our study paves the way for a deeper understanding of human abstraction learning and generalization.