Introduction

Fused nitrogen heterocyclic compounds have received increasing attention in recent decades due to their significant value in the field of medicine1,2. The physicochemical properties of these compounds are significantly influenced by the type and size of the ring structure, as well as the substituent groups on the core skeleton3. Particularly, quinazolinones and purines, as key structural motifs in many natural products and bioactive compounds4,5, often exhibit novel physicochemical or biological properties when added with an additional ring6. Studies have shown that 2D tricyclic derivatives exhibit enhanced fluorescence properties compared to their 2D bicyclic parent compounds, but with reduced bioactivity (Fig. 1a)7,8,9. In contrast, their 3D tricyclic derivatives exhibit superior pharmacokinetic properties and pharmacological activity, in line with the “escape from flatland” strategy (Fig. 1a)10,11,12,13. Furthermore, structure-activity relationship studies have confirmed that optically pure compounds possess significantly enhanced bioactivity compared to their racemic counterparts14,15. Consequently, the synthesis of chiral semisaturated aza-heterocycles has remained a hot topic in organic synthesis. Traditional synthetic methods rely on cyclization and annulation reactions10,16,17, which often involve multi-step reaction sequences, pre-functionalized starting materials, and/or harsh reaction conditions. Direct dearomatization of heteroaromatics is an efficient strategy for constructing semisaturated aza-heterocycles, but the reported methods primarily involve electron-rich indoles18. Methods for directly converting electron-deficient aza-fused heterocycles into chiral semisaturated aza-heterocycles are rarely reported19,20.

Fig. 1: Development of stereoselective skeleton editing of polynitrogen heterocycles.
Fig. 1: Development of stereoselective skeleton editing of polynitrogen heterocycles.The alternative text for this image may have been generated using AI.
Full size image

a Advantages of semi-saturated tricyclic nucleosides in pharmacokinetic properties or biological properties. b Ring contraction reactions: from 3D aliphatic six-membered ring to 3D aliphatic five-membered ring; from 2D aromatic six-membered ring to 2D aromatic five-membered ring. c Design and challenges of stereoselective ring contraction from a 2D six-membered ring to a 3D five-membered ring. d This work: stereoselective skeletal editing of fused polyazapolycycles.

The skeletal editing strategy is designed to generate reactive intermediates by disrupting the inherent molecular framework, followed by reconfiguration, thereby facilitating swift and facile alterations to the core structure21,22,23,24,25,26. This approach serves as a potent instrument for the rapid and efficient development of drug candidates within the field of medicinal chemistry27,28,29,30,31,32,33,34. Among these, the transformation of a three-dimensional (3D) aliphatic six-membered ring into a 3D aliphatic five-membered ring is particularly well-advanced (Fig. 1b)35,36,37,38. Notably, the photoinduced ring-rearrangement of saturated heterocycles, as recently reported by Sarpong’s group, has marked a significant advancement in this area39,40. The strategies for skeletal editing of two-dimensional (2D) aromatic six-membered rings to 2D aromatic five-membered rings have also been rapidly evolving (Fig. 1b)21,35. Harran41,42, Zheng43, and Fu44 have reported innovative methods for the transformation of pyridine structures into pyrrole and pyrazole frameworks, respectively. Levin’s group also introduced a novel approach to prepare indole compounds by removing carbon from quinoline skeleton through photochemical reaction45,46. Furthermore, McNally47, Sarpong48 and Fu49 have independently reported effective strategies for the synthesis of pyrazole and oxazole scaffolds through the deconstruction-reconstruction of pyrimidine skeletons. Despite these studies’ significant contributions to scientific progress in the field, asymmetric skeletal editing methods for aza-heterocycles are limited, with only You50 and Bi’s51 groups reporting asymmetric annulation reactions of indoles. Selective editing of pyrimidine-fused aza-heterocycles, particularly the stereoselective ring contraction rearrangement of 2D aza-heterocycles to chiral 3D aza-heterocycles, remains an underexplored area.

Chiral hypervalent iodine reagents are pivotal tools in organic synthesis, increasingly playing a significant role in the field of asymmetric synthesis52. Recent research on chiral hypervalent iodine reagents has made remarkable progress, particularly in applications such as asymmetric dearomatization, α-functionalization of carbonyls, difunctionalization of alkenes, alkene oxidation rearrangements, and oxidative coupling reactions52,53. However, the electron-deficient nature, high polarity, poor solubility of pyrimidine-fused heterocycles, and the presence of multiple reactive sites54,55,56 pose significant challenges to the high chemo-, regio-, and enantioselective ring contraction rearrangement of pyrimidine catalyzed by chiral hypervalent iodine. Building on our previous work with hypervalent iodine mediated dialkoxylation of electron-rich indole compounds57, we hypothesized that the introduction of a methoxy group into the electron-deficient pyrimidine ring, followed by rearrangement with hypervalent iodine, could be a viable approach (Fig. 1c). The in situ introduction of the methoxy group not only alters the electronic properties of the compound but also enhances its solubility. Moreover, the methoxy group is an important pharmacophore group that plays a significant role in ligand-target interactions, the physicochemical properties of drugs, and ADME (absorption, distribution, metabolism, and excretion) characteristics58. We herein report a chiral hypervalent iodine(III)-catalyzed ring contraction rearrangement of pyrimidine moieties to form multi-substituted imidazolidine rings, successfully achieving a highly chemo-, regio- and stereo-selective transformation from 2D nitrogen heterocycles to fused 2D/3D nitrogen heterocycles (Fig. 1d). Compared with existing methods for imidazoline synthesis59,60,61, this approach exhibits superior atom economy, step economy, and substrate versatility.

Results and discussion

Reaction design and optimization

To initiate this study, tricyclic quinazolone 1a was selected as a model substrate for investigating the asymmetric ring-contraction reaction catalyzed by chiral hypervalent iodine. The utilization of chiral iodoarene 3a as a precatalyst and MeOH as both reagent and solvent did not lead desired result (Fig. 2, entry 1). Consequently, alternative chiral iodoarene precursors with distinct backbones were explored. Encouragingly, when chiral iodoarene 4a was employed, the desired compound 2a was obtained with excellent stereoselectivity, albeit a poor enantioselectivity, yielding 66% (Fig. 2, entry 3). To enhance the reaction outcome, the effect of acid additives was investigated (Fig. 2, entries 5-8). Employing TfOH resulted in an increased yield of compound 2a, reaching 84%. Notably, replacing MeOH with DCM as the solvent further enhanced the enantioselectivity to 56% ee (Fig. 2, entry 9). Subsequently, additional investigations focused on exploring the effectiveness of different chiral iodoarenes. Using the (S)-proline-derived chiral organoiodine 5c, the desired compound 2a was obtained with excellent stereoselectivity, exhibiting a yield of 89% and an impressive enantioselectivity of 97% ee (Fig. 2, entry 13). Regrettably, the expansion to other nucleophiles did not yield the expected results. The addition of benzylamine triggered the decomposition of starting material 1a, while nucleophiles such as acetic acid, thiophenol, and methyl acetoacetate failed to induce rearrangement, with most of 1a being recovered (see ESI, Table S5).

Fig. 2: Optimization of the reaction conditions.
Fig. 2: Optimization of the reaction conditions.The alternative text for this image may have been generated using AI.
Full size image

Reaction conditions: 1 (0.1 mmol, 1.0 equiv), MeOH (0.4 mL, 99.0 equiv), chiral ArI (20 mol%), selectfluor (2.0 equiv), acid (3.0 equiv), solvent (4.0 mL) at 30 °C of x h. Yield of isolated product. The enantiomeric excess (ee) values were determined by HPLC. The diastereomeric ratio (d.r.) was determined by 1H NMR spectroscopy of the crude reaction mixture.

Substrate scope of pyrimidine fused heterocycles and alcohols

With the optimized reaction conditions established, the substrate scopes and limitations of this protocol were explored (Fig. 3). The quinazolinone bearing different groups on the phenyl ring afforded the polysubstituted fused-ring guanidines (2a-2l) with moderate to good yields (35-89%), excellent enantioselectivities (91-99% ee) and diastereoselectivities (>20:1 d.r.). 2a was synthesised in 73% yield and 98% ee in a 1 mmol scale reaction. The absolute configuration of products 2a and 2 f was unambiguously confirmed by single-crystal X-ray diffraction. The polyfluoro-substituted and fused tetracyclic substrates also reacted smoothly, and the ee values of the products 2j-2l were all above 90%. Subsequently, using primary alcohols (MeOH-d4, ethanol and cyclopropylmethanol) as nucleophiles, the desired products 6a-6c were obtained in 52-80% yields with up to 97% ee. In addition, the versatility of this synthetic approach for nitrogen-containing bicyclic substrates was investigated. Under standard reaction conditions, two pyrimidine-fused pyrimidines 7a and 7b were successfully converted to pyrimidine-fused imidazoline products 8a and 8b. The methodology was extended to the synthesis of 5,5-fused bicyclic systems, in particular using triazole-fused pyrimidine 7 d to produce triazole-fused imidazoline 8 d in moderate yields and excellent enantioselectivity (93% ee).

Fig. 3: Substrate scope of fused polyazapolycycles and alcohols.
Fig. 3: Substrate scope of fused polyazapolycycles and alcohols.The alternative text for this image may have been generated using AI.
Full size image

Reaction conditions: 1 or 7 (0.1 mmol, 1.0 equiv), 5c (20 mol%), selectfluor (2.0 equiv), TfOH (3.0 equiv), co-solvent DCM:ROH (v/v 10:1, 4.4 mL) at 30 °C. Yield of isolated product. The ee values were determined by HPLC. The diastereomeric ratio (d.r.) was determined by 1H NMR spectroscopy of the crude reaction mixture. aDCM:MeOH (v/v 3:2, 4.4 mL) was used as solvent. bDCM:ROH (v/v 1:1, 4.4 mL) was used as solvent.

Application in late-stage functionalization

Encouraged by the aforementioned experimental results, we intend to apply the developed method to the late-stage functionalization of endogenous nucleoside M1G-dR and its derivatives (Fig. 4). Initially, alkyl-substituted tricyclic nucleosides (ethyl, n-butyl and benzyl) were used as substrates, giving the target products 10a-10c in 63-65% yields and 93-95% ee values. Acyclovir, an antiviral drug against herpes simplex virus, and its series of amino acid and peptide-linked derivatives react with TMOP to form M1G-like acyclic nucleosides62. These substrates also underwent selective skeletal editing of the pyrimidine ring under standard conditions, yielding the annulated products 10d-10i with good yields, high diastereoselectivity and excellent enantioselectivity (>90% ee). The method developed can be directly applied to the late-stage functionalization of unprotected M1G-dR 9j and M1G-R 9 l, albeit with less than ideal yields. We have tried changing different chiral columns and mobile phases, as well as derivatising the product, but failed to separate the diastereoisomers by HPLC. Under the standard reaction conditions, the Obz-protected derivatives of M1G-dR 9k and M1G-R 9 m exhibited significantly enhanced reactivity, yielding the desired products 10k and 10 m with yields of 28% and 67%, respectively. The tricyclic guanosine derivative with a 5’-position azide substitution also reacted to form the corresponding target product 10n, with the azide group serving as a reactive site for further structural modification. In nature, nucleosides require activation by kinases to form nucleotides, and modified nucleosides are challenging to phosphorylate in vivo, thus making the late-stage functionalization of nucleotides highly significant. Using the tricyclic nucleotides 9o and 9p as substrates, the target products 10o and 10p were synthesized under standard conditions. Notably, when the substrates were M1G-containing oligonucleotides, the desired products 10q and 10r were obtained with yields of 18% and 57%, respectively. Compared with their acyclic counterparts, nucleoside, nucleotide, and oligonucleotide substrates performed poorly under identical conditions. We attribute this outcome to the lability of the glycosidic bond, which leads to partial decomposition of the starting material. Attempts to optimize the reaction parameters provided no improvement.

Fig. 4: Late-stage functionalization of M1G-dR and its analogues.
Fig. 4: Late-stage functionalization of M1G-dR and its analogues.The alternative text for this image may have been generated using AI.
Full size image

Reaction conditions: 9 (0.1 mmol, 1.0 equiv), 5c (20 mol%), selectfluor (2.0 equiv), TfOH (3.0 equiv), co-solvent DCM:MeOH (v/v 1:1, 4.4 mL) at 30 °C. Yield of isolated product. The ee values were determined by HPLC. The diastereomeric ratio (d.r.) was determined by 1H NMR spectroscopy of the crude reaction mixture. aKOH (3.0 equiv) was used as an additive. bThe ee values were determined by HPLC from the hydrolysed product. See Supplementary Information for details.

To showcase the synthetic utility of the current methodology, the recycling of chiral catalysts and derivatization reactions of the tricyclic nucleosides were performed (Fig. 5). The chiral catalyst re-5c was reused and recycled three times with minimal change in the yield and enantiomeric excess of the target product 2a (Fig. 5a). Compounds 10k and 10 m were deprotected to give 10j and 10 l in excellent yields. The azide-substituted product 10n underwent a click reaction with 6-bromo-9-(prop-2-yn-1-yl)−9H-purine to afford the triazole-linked oligonucleotide derivative 11 in a 92% yield (Fig. 5b).

Fig. 5: Synthetic potentials.
Fig. 5: Synthetic potentials.The alternative text for this image may have been generated using AI.
Full size image

a The catalyst recovery and recycling experiments. The scale-up synthesis. b The derivatization reaction of tricyclic nucleosides.

Mechanistic considerations

Subsequently, we conducted a preliminary mechanistic investigation of the rearrangement reaction (Fig. 6). Under the standard conditions, pyrido[1,2-a]quinoxaline 12 failed to react, underscoring the pivotal role of the N-1 atom in 1a in initiating the reaction (Fig. 6a, eq 1). When m-CPBA or PhI(OAc)₂ were employed as oxidants in place of Selectfluor, m-CPBA afforded the product 2a in 46% yield with 98% ee, whereas PhI(OAc)₂ delivered the product 2a in excellent yield but as a racemate (91% yield, 0% ee). These results suggest that Selectfluor first oxidizes the chiral aryl iodide 5c to a hypervalent iodine species, and this newly formed catalyst then drives the rearrangement reaction (Fig. 6a, eq 2 and 3). When PhI(OMe)₂ was employed directly as the catalyst, rac-2a could be obtained in 55% yield, indicating that the in-situ-generated Ar*I(OMe)₂ is the active catalytic species (Fig. 6a, eq 4).

Fig. 6: Mechanistic observations.
Fig. 6: Mechanistic observations.The alternative text for this image may have been generated using AI.
Full size image

a Control experiments. b Calculated potential energy profile for Ar*I(OMe)2-catalyzed stereoselective skeletal editing of 1a. See Supplementary Data 1 for the cartesian coordinates of the optimized structures. c Calculated geometries of transition structures and intermediates.

Guided by literature precedents63,64 and the above experimental findings, we elucidated the mechanism by means of DFT calculations (Fig. 6b). First, substrate 1a adds to the in-situ-generated chiral hypervalent iodine (Ar*I(OMe)2) through TS1 to give Int1. The transition state TS1 is 9.2 kcal/mol lower in energy than the TS1’, indicating that the S-configured Int1 is favored (Fig. 6c). Ligand exchange with TfOH then converts Int1 into Int2. In Int2, the methoxy group attached to the hydrogenated pyrimidine moiety significantly stabilizes the iodine(III) cation, with an energy 5.5 kcal/mol lower than that of Int2’. A syn-addition (TS2) of the MeO- and the hypervalent iodine (Ar*IOMe) across Int2 furnishes Int3, consistent with Ariafard’s observation of facile syn-additions of phenols with MeOH and hypervalent iodine reagents65. The low barrier of 5.0 kcal/mol for this step dictates the high chemo-, regio-, and stereoselectivity observed. Int3 subsequently releases MeO- to afford Int4. The hydropyrimidine unit in Int4 has three cis-oriented substituents that cause severe steric hindrance, resulting in a highly distorted conformation. Upon rearrangement (TS3), this motif transforms into an imidazoline ring. C-2 migrates to an exocyclic position, resulting in the loss of stereochemistry. Meanwhile, C-3 undergoes an intramolecular nucleophilic substitution that inverts its configuration, delivering the trans-configured Int5 and releasing Ar*I (4c). Finally, Int5 reacts with MeO- to deliver the product 2a.

In conclusion, we have developed a highly chem-regio- and stereoselectivity method for the construction of polysubstituted 3D fused-ring guanidines via a chiral hypervalent iodine(III)-catalyzed ring-contraction rearrangement reaction of pyrimidine rings within polynitrogen heterocycles. The application of this method to the late-stage skeletal modification of endogenous nucleoside M1G-dR and its analogues involving serial polyatom changes is demonstrated. To demonstrate its versatility, the scaled-up synthesis, recycling of chiral catalysts and derivatization reactions of the tricyclic nucleoside derivatives were successfully carried out.

Methods

General procedure for the synthesis of chiral products

General procedure A

Nitrogen-containing heterocycle (0.1 mmol, 1.0 equiv), 5c (0.02 mmol, 20 mol%), selectfluor (0.2 mmol, 2.0 equiv) and DCM:ROH (v/v 10:1, 4.4 mL) were added to a screw-capped vial with a magnetic stirrer. TfOH (0.3 mmol, 3.0 equiv) was then rapidly added to the reaction mixture. The vial was tightly sealed and stirred at 30 °C for 3-24 h. The reaction was diluted with DCM and washed with a saturated aqueous solution of NaHCO3. Extract the aqueous phase with DCM. The combined organic phases were dried over anhydrous Na2SO4 and concentrated under reduced pressure. The residue was subjected to silica gel chromatography to give target compounds. The enantiomeric excess was determined by HPLC with a chiral column.

General procedure B

KOH (0.3 mmol, 3.0 equiv), 5c (0.02 mmol, 20 mol%), selectfluor (0.2 mmol, 2.0 equiv) and DCM:MeOH (v/v 1:1, 4.4 mL) were added to a screw-capped vial with a magnetic stirrer. TfOH (0.3 mmol, 3.0 equiv) was then rapidly added to the reaction mixture followed by tricyclic nucleoside (0.1 mmol, 1.0 equiv). The vial was tightly sealed and stirred at 30 °C for 12 h. The reaction was diluted with DCM and washed with a saturated aqueous solution of NaHCO3. Extract the aqueous phase with DCM. The combined organic phases were dried over anhydrous Na2SO4 and concentrated under reduced pressure. The residue was subjected to silica gel chromatography to give target compounds. The enantiomeric excess was determined by HPLC with a chiral column.

General procedure C

Tricyclic nucleoside (0.1 mmol, 1.0 equiv), 5c (0.02 mmol, 20 mol%), selectfluor (0.2 mmol, 2.0 equiv) and DCM:MeOH (v/v 1:1, 4.4 mL) were added to a screw-capped vial with a magnetic stirrer. TfOH (0.3 mmol, 3.0 equiv) was then rapidly added to the reaction mixture. The vial was tightly sealed and stirred at 30 °C for 12 h. The reaction was diluted with DCM and washed with a saturated aqueous solution of NaHCO3. Extract the aqueous phase with DCM. The combined organic phases were dried over anhydrous Na2SO4 and concentrated under reduced pressure. The residue was subjected to silica gel chromatography to give target compounds. The enantiomeric excess was determined by HPLC with a chiral column.