Carbonyl swapping converts cyclic ketones to saturated heterocycles

Abstract

Saturated heterocycles are privileged scaffolds in bioactive molecules 1 . Despite the availability of numerous de novo routes to various heterocyclic compounds, accessing diverse heterocycles from a unified, readily available carbocycle would offer a strategic alternative for constructing challenging heterocyclic structures from unconventional precursors 2 . Here we report a modular approach that transforms a single cyclic ketone into various saturated heterocycles through formal carbonyl replacement with heteroatoms, by a scarcely explored bis(aroylperoxy) ketal intermediate. Through electronically guided peroxy cleavage, this intermediate enables double C–C bond scission of cyclic ketones, generating alkyl dichlorides as versatile handles for modular N/O/S/Se/Te incorporation using simple nucleophiles. This method exhibits broad substrate scope and functional-group tolerance, enabling both accelerated target synthesis and late-stage diversification of bioactive molecules. Its utility is also extended through ‘ring construction–carbonyl replacement’ and ‘ring functionalization–carbonyl replacement’ strategies, in which cyclic ketones prepared by well-established methods are converted into challenging-to-access heterocycles for which analogous methods remain underdeveloped. By combining C–H oxidation with carbonyl replacement, a proof-of-concept formal ‘CH 2 -to-heteroatom’ conversion is further demonstrated.

Main

Heterocycles are ubiquitous in drug molecules and profoundly influence their pharmacological properties 3 , as exemplified by the marked differences that even single-atom variation in heterocycle identity may produce in otherwise analogous compounds 4 , 5 , 6 , 7 , 8 (Fig. 1a). Although numerous methods exist for the de novo synthesis of heterocyclic compounds 9 , strategies that convert pre-existing carbocycles into saturated heterocycles remain limited and often constrained by specialized substrate requirements, customized routes and/or the restricted types of heterocycles accessible 10 . Among these, a particularly attractive strategy would be through single-atom replacement 11 , 12 , 13 , 14 , 15 , 16 , which could, in principle, enable the programmable replacement of a ring carbon atom with various heteroatoms, thus offering divergent access to diverse saturated heterocycles from a unified carbocycle with considerable retrosynthetic simplicity 17 . This transformation, sometimes referred to as a type of ‘ring transmutation’ 18 , would provide expanded opportunities to access challenging heterocyclic structures from unconventional precursors and enable skeletal modification of saturated rings embedded in late-stage intermediates and/or natural scaffolds 19 , 20 . Moreover, it may facilitate more efficient access to diverse arrays of heterocyclic analogues, thus aiding structure–activity relationship studies 21 . Yet, despite considerable recent advances in the field of skeletal editing 18 , 22 , 23 , 24 , 25 , 26 , 27 , this vision remains largely unrealized for saturated ring scaffolds, particularly given the wide range of second-, third-, fourth-, and even fifth-row heteroatoms that are chemically feasible for this in-ring replacement. To date, most efforts in the field have focused on unsaturated or partially unsaturated scaffolds 11 , 12 , 13 , but protocols for single-atom replacement in saturated rings—particularly those enabling access to diverse heterocycles 16 —have received far less attention 14 , 15 , 16 .

a, Ubiquity of saturated heterocycles in bioactive molecules and their profound influence on pharmacological properties. b, This work: converting cyclic ketones to diverse heterocycles by modular carbonyl-to-heteroatom swap. Me, methyl; IC 50 , half-maximal inhibitory concentration; ICMT, isoprenylcysteine carboxyl methyltransferase; TMS, trimethylsilyl; equiv., equivalent; r.t., room temperature.

Additionally, the practical utility of these transformations will also depend on the precursor ring structure, as an ideal candidate should be highly prevalent, structurally diverse and readily available through synthesis (Extended Data Fig. 1a). In this regard, cyclic ketones emerge as a compelling choice, first owing to their superior synthetic accessibility among common saturated cyclic scaffolds, particularly evidenced by the numerous established methods available for their ring construction and elaboration 10 . Moreover, the ubiquitous occurrence of cyclic ketones and their common precursors (for example, secondary alcohols, enones) in natural products further underscores their strategic advantages as starting ring scaffolds 28 . However, developing this divergent ring transmutation of cyclic ketones still faces notable challenges, particularly the consecutive cleavage of two strong and kinetically inert C–C bonds adjacent to the carbonyl group. Moreover, to achieve high modularity for versatile carbonyl-to-heteroatom (CO-to-X) swaps, the transformation would further benefit from both unified reaction protocols and the use of readily available, native heteroatom sources that require no pre-functionalization. Although many strategies have been successfully developed for single C–C bond activation of ketones (for example, by O- and N-centred radicals or aromatization-driven strategies 29 , 30 , 31 , 32 ), methods enabling sequential double C–C cleavage of ketones remain scarce 33 (Extended Data Fig. 1b). A recent study 15 achieved an elegant—yet specific—carbonyl-to-sulfur swap through a relatively synthetically demanding ketone-activating reagent and a Ts 2 S dual-radical acceptor, illustrating both the promise and the current limitations of the field. Here, we report the divergent transformation of cyclic ketones into diverse heterocycles, proceeding through a unique bis(aroylperoxy) ketal intermediate (Fig. 1b). This species is isolable, prepared from ketones at low cost and can be used without purification on a practical scale. Through sequential release of two oxygen-centred radicals anchored to the ketone framework, this intermediate enables two consecutive C–C bond cleavages and two concomitant C–halogen bond formations, thereby providing a general platform for modular CO-to-N/O/S/Se/Te swap with diverse nucleophiles.

Our design is inspired by the well-documented ability of oxygen-centred radicals to cleave unactivated C–C bonds through β-scission 34 , 35 , 36 , 37 , and their regio-predictable generation from peroxides through single-electron reduction, with spin centres localizing on the less electronegative oxygen fragments 36 (Extended Data Fig. 2a). In the context of ketone C–C cleavage, a well-explored approach involves converting the ketone into a hydroperoxy hemiketal (1), which acts as a precursor to a geminal hydroxy oxyl radical, enabling a single C–C cleavage 30 , 37 . Featuring the installation of two peroxy units onto a ketone, bis(hydroperoxy) ketals (2) have also been recently investigated, yet these structures still fail to deliver two productive oxygen-centred radicals able to induce C–C cleavage 38 . Only the first reductive cleavage of the initial peroxy unit in 2 generates an oxygen radical that triggers a C–C cleavage. The second reductive cleavage event on the resulting peracid intermediate yields at best a hydroxyl radical and a carboxylate anion—neither of which promotes the second C–C cleavage. To address this, we proposed further attaching two strongly electron-withdrawing groups (EWGs) to the terminal oxygens of the bis(hydroperoxy) ketals, leading to structure 3. By imposing a strong electronic bias, these EWGs are expected to orchestrate the reductive cleavage of both peroxy bonds to sequentially release two oxygen radicals tethered to the carbon framework, thus making 3 a synthetic equivalent of gem-dioxyl diradicals, which can break both α C–C bonds of the ketone (Extended Data Fig. 2b).

Accordingly, a more detailed working mechanism is proposed, centred on bis(acylperoxy) ketal (5) as a representative example of 3—a scaffold scarcely explored in organic synthesis yet accessible by a perhydrolysis–acylation sequence from ketones (4) (refs. 39 , 40 , 41 , 42 ) (Fig. 2a). Treatment of 5 with an Fe(II) catalyst induces single electron transfer (SET), generating an oxygen radical (Int 1) that triggers the first C–C cleavage. The resulting alkyl radical (Int 2) undergoes halogen atom transfer (XAT) with an Fe–Cl species 43 , 44 , forming the first C–Cl bond and giving a mixed diacyl peroxide (Int 3). A second round of Fe-mediated SET/XAT then triggers a second C–C cleavage, yielding dichloride 6, which can react with a range of subsequently added nucleophiles to furnish diverse heterocycles in a modular fashion. During reaction optimization, the electron-withdrawing ability of the acyl group further proved important, with more electron-deficient variants giving higher yields (for example, of 6a) and suppressed by-product formation (for example, 8; Fig. 2b). This is probably because the second SET event from the mixed diacyl peroxide (Int 3) can, in principle, produce either of two carboxyl radicals, and the relative electron-withdrawing strengths of the acyl groups (that is, the degree to which they stabilize the incipient carboxylate anion) determine the proportion of the productive carboxyl radical (that is, Int 4) formed 45 . Ultimately, the 3,5-bis(trifluoromethyl)benzoyl substituent emerged as optimal, providing both high conversion and selectivity. Further fine-tuning of the reaction parameters established the optimized conditions for the CO-to-N replacement (Fig. 2c). Starting from 4, crude 5c is generated and carried into the activation stage after a simple silica gel pad filtration. Treatment of the resultant dichloride with benzylamine then affords the desired CO-to-N exchange product 7a. Note that the entire procedure requires only a single chromatographic purification (of 7a), delivering an overall yield of 86% on a 1 mmol scale. In an alternative workflow, 5c can be isolated by column chromatography on silica gel for full characterization; subjecting this purified material to the second stage then gives 7a in a similar yield.

a, Detailed mechanism of the overall transformation. b, Effect of the electron-withdrawing strength of acyl. c, Optimized protocol for CO-to-N replacement using 5c. Yields were reported as 1 H NMR yields for Fig. 2b and as isolated yields for Fig. 2c. For details, see Supplementary Information section 3. SET, single electron transfer; XAT, halogen atom transfer; Int, intermediate; Ph, phenyl; TMS, trimethylsilyl; Et, ethyl; equiv., equivalent; r.t., room temperature; Bn, benzyl.

With the optimized protocols in hand, we next explored the substrate scope of the transformation (Fig. 3). A portion of the ketone substrates were subjected to the telescoped procedure without intermediate purification, and for others, the bis(aroylperoxy) ketal intermediates (typically prepared in >80% yields on gram scales) were isolated and characterized to unambiguously verify their intermediacy in the entire transformation. To our delight, the reaction proved broadly tolerant of diverse functional groups (FGs), including esters (29), protected alcohols (30), amides (31), sulfonamides (35), ethers (38), pyridines (16), aryl chlorides, bromides (14) and iodides (15). More sensitive functionalities such as alkenes (36), unprotected tertiary alcohols (37) and lactones (40) were also found to be compatible. The transformation is not restricted to six-membered rings: five-, six- and seven-membered cyclic ketones, both monocyclic and benzo-fused, underwent efficient CO-to-N replacement in moderate to good yields. The protocol can be extended to more structurally complex substrates to deliver valuable heterocyclic building blocks such as polysubstituted (44), spirocyclic (40) and cyclopropane-containing piperidines (42), and its utility was further demonstrated by the facile remodelling of steroidal frameworks into aza-steroid derivatives (46 and 48), again highlighting its potential in late-stage skeletal modification and medicinal chemistry applications. This transformation also accommodates a wide range of amine nucleophiles, including both aliphatic amines and anilines. Through the formal extrusion of a CO unit from a cyclic ketone and insertion of an amine motif, it offers a unique and complementary retrosynthetic disconnection to access medicinally privileged cyclic amine scaffolds, particularly those not easily accessible using existing methods. For instance, cyclic amines bearing bulky α-tertiary alkyl groups, such as adamantyls (65 and 66) and bicyclopropyls (60), or transition-metal-sensitive FGs such as aryl boronates (63) and iodides (64), can now be readily prepared from feedstock chemicals using this approach. These structures are challenging to obtain by reacting cyclic secondary amines with electrophiles by conventional S N 2 or C–N coupling reactions 46 . Other FGs commonly found in drug molecules, such as free hydroxy groups (66), Boc-protected amines (60), cyclopropyl (62) and gem-difluorocyclobutyl moieties (61), are also well tolerated and can be conveniently introduced. Moreover, this ‘pop-and-dope’ approach also enables straightforward access to a variety of high-value chiral cyclic amines from readily available chiral cyclic ketones. As the transformation does not perturb stereogenic centres on the primary amine nucleophiles, it permits the formal fusion of chiral ketones with chiral primary amines—both among the most accessible chiral building blocks 47 —to modularly assemble complex, drug-like tertiary amines bearing multiple stereocentres with exclusive stereospecificity (70, 72, 74 and 76). This reaffirms the unique strategic value of ring transmutation methods targeting saturated carbocycles.

Reactions performed on a 0.2–1 mmol scale. Isolated yields reported. a The bis(aroylperoxy) ketal intermediate was synthesized using the ethylene glycol ketal that was carried over from starting material preparation. b Using 100 mol% FeCl 3 . c C–N bond formation performed with Cs 2 CO 3 , KI, DMSO at 80 °C. For details, see Supplementary Information section 4. L.A., Lewis acid; cat., catalyst; Et, ethyl; r.t., room temperature; TMS, trimethylsilyl; equiv., equivalent; Bn, benzyl; TBS, tert-butyldimethylsilyl; Cbz, benzyloxycarbonyl; Ar, 3,4-difluorophenyl; Ts, para-toluenesulfonyl; Ph, phenyl; Boc, tert-butoxycarbonyl; Bpin, pinacol boronate; e.r., enantiomeric ratio.

To unlock broader potential of this modular ring-editing strategy, we next explored the transformation of cyclic ketones to diverse heterocycles by using a similar reaction sequence but diverging at the final stage through the choice of different heteroatom nucleophiles (Fig. 4). After optimization, we were delighted to find that a variety of readily available bis-nucleophiles—such as Na 2 S, NaOH, NaHSe, Na 2 Te or carbon nucleophiles bearing two acidic C–H bonds—can all be incorporated under tailored conditions, leading to cyclic ethers, thioethers, selenides, tellurides or fully saturated carbocycles derived from cyclic ketones in a divergent manner. Note that the bis(aroylperoxy) ketals can be shared as common intermediates across these transformations, thereby further reducing the overall step count for accessing diverse heteroatom-substituted analogues. For example, applying carbonyl-to-heteroatom replacement to the A-ring of 5α-androstane-3,17-dione (77) efficiently generated a series of oxa-, thia-, selena- and tellura-steroid derivatives—valuable yet underexplored compounds that are non-trivial to prepare by conventional approaches 48 . In another set of experiments, the CO-to-X swap was used to transform inexpensive, ketone-based chiral building blocks into diverse, challenging-to-access chiral hetero- and carbocycles (for example, 83–86). This is particularly highlighted by the straightforward conversion of commercial chiral piperidone (5 USD g −1 , reaxys) into 15 N-labelled chiral piperazine 86 and chiral thiomorpholine 85; the latter skeleton previously required more than five steps to prepare from chiral 1,2-amino alcohols 49 . This protocol was further applied to biologically active molecules. For example, 87, an α-adrenergic receptor antagonist 50 , could be readily converted into a panel of single-atom ring-modified analogues by modular CO-to-N/O/S/Se/Te replacement. Similarly, facile CO-to-N/S exchange of a protected ketone (93) provided not only compound 94 as a human H 3 receptor antagonist 51 but also a unified synthetic route to other heterocyclic analogues of potential interest. Similar protocols were also exemplified using a liquid-crystal molecule (97).

Reactions performed on a 0.2 mmol scale except for 87. Isolated yields are reported. a Using 100 mol% FeCl 3 . For details, see Supplementary Information section 5. L.A., Lewis acid; cat., catalyst; TBA, tetrabutylammonium; DMF, N,N-dimethylformamide; EWG, electron-withdrawing group; NaHMDS, sodium hexamethyldisilazide; 15-C-5, 15-crown-5; 2-MeTHF, 2-methyltetrahydrofuran; e.r., enantiomeric ratio; d.r., diastereomeric ratio; Ts, para-toluenesulfonyl; PMP, para-methoxyphenyl; i Pr, isopropyl; SM, starting material.

To further highlight the synthetic utility of the established CO-to-X replacement, we moved forward to demonstrate its application in the ‘ring construction–carbonyl replacement’ and ‘ring functionalization–carbonyl replacement’ strategies (Fig. 5). This is done by first leveraging the rich and mature chemistry available for the assembly or functionalization of a cyclic ketone scaffold—such as cycloaddition reactions, enolate chemistry and transition-metal catalysed processes—followed by the on-demand replacement of the carbonyl group with a desired heteroatom. These strategies not only offer rapid access to heterocycles that otherwise require multistep syntheses but also provide valuable retrosynthetic simplicity 17 , as many heterocycles can now be traced back to cyclic ketones as new, versatile precursors. For example, by conducting the classic [2 + 2 + 1] cyclopentanone synthesis using Pauson–Khand-type reactions, followed by the sequential carbonyl replacement with heteroatoms, one can establish a general route to access diverse 2,5-diheteroatom-substituted bicyclo[3.3.0]octanes (103–105) as valuable heterocyclic building blocks. Similarly, a ring-construction phase involving a three-component Mannich reaction, followed by CO-to-N/S swaps, provides direct access to 3-thia-8-azabicyclo[3.2.1]octanes (107) and 3,8-diazabicyclo[3.2.1]octanes (108), whose prior de novo syntheses required nine steps each and diverged at an early stage 52 . Moreover, Diels–Alder reactions between 2-siloxy-1,3-butadienes and dienophiles are widely applied to construct densely substituted cyclohexanones with precise stereocontrol (for example, 109). These can now be combined with subsequent carbonyl replacement to give densely substituted heterocycles (for example, 110 and 111) with the same stereochemical precision. Finally, an example for the ring functionalization–carbonyl replacement strategy is also demonstrated using ketone 112. Using a recently developed enzymatic enantioselective desaturation 53 followed by Pd-catalysed conjugate addition of the resulting enone, trisubstituted ketone 113 bearing two contiguous stereocentres was readily prepared in two steps. Subsequent CO-to-N exchange then furnished the corresponding chiral piperidine 114 in 55% yield with complete stereospecificity. Taken together, we believe these examples highlight a unique synthetic opportunity enabled by our method: the opportunity to leverage the extensive synthetic toolkit available for a specific ring system (that is, cyclic ketones) to access diverse heterocycles for which analogous methods have remained elusive.

Isolated yields are reported. a Reference 53 . Reaction condition: GkOYE-C26S-D73G, FMN, air, buffer, 50 °C. b Pd(OAc) 2 , TfOH, NaNO 3 , DCE, 40 °C. c The minor diastereomer was separated and discarded at the bis(aroylperoxy) ketal stage. For details, see Supplementary Information section 6. Ts, para-toluenesulfonyl; i Pr, isopropyl; Bn, benzyl; Ac, acetyl; Cbz, benzyloxycarbonyl; Ph, phenyl; TMS, trimethylsilyl; Tf, triflyl; GkOYE, old yellow enzyme from Geobacillus kaustophilus; FMN, flavin mononucleotide; e.r., enantiomeric ratio; d.r., diastereomeric ratio; DCE, 1,2-dichloroethane.

Finally, we envision that our carbonyl-to-heteroatom replacement might be further integrated with site-selective C–H oxidation—particularly of the methylene-to-carbonyl type—to achieve a two-phase methylene-to-heteroatom (CH 2 -to-X) replacement by a transient oxo group. Given that chemical or enzyme-catalysed C–H oxidation is among the most feasible approaches for functionalizing unactivated methylene sites 54 , with ever-growing diversity in achievable site-selectivity, such an overall CH 2 -to-X replacement holds great promise for selectively knocking off unactivated, skeletal CH 2 units from molecules of interest and replacing them with desired heteroatoms. Although a full exploration of this transformation is apparently beyond the main focus of this study, a preliminary proof of concept is nevertheless demonstrated. As shown in Fig. 6, subjecting 115 to White’s C–H oxidation conditions 55 furnishes ketone 116, which is then converted by our CO-to-X replacement into the corresponding N- and S-incorporated structures (117 and 118), respectively. Beyond this intriguing direction, further improvement of the divergent ring transmutation protocol for cyclic ketones, as well as exploration of bis(aroylperoxy) ketal intermediates for other unconventional transformations, is currently underway in our laboratory and will be reported in due course.

Isolated yields are reported. a Reference 55 . Reaction condition: (1) MeOTf, DCM; (2) Fe(CF 3 PDP), AcOH, H 2 O 2 , MeCN; then demethylation. For details, see Supplementary Information section 7. Tf, triflyl; Fe(CF 3 PDP), bis(acetonitrile)[(2S,2′S)-1,1′-bis((5-(2,6-bis(trifluoromethyl)phenyl)pyridin-2-yl)methyl)-2,2′-bipyrrolidine]iron hexafluoroantimonate; AcOH, acetic acid; Me, methyl; Bn, benzyl; d.r., diastereomeric ratio.