Introduction

N-glycosylated heterocycles are important in biological systems and drug development. For instance, N-glycosylated (ribosylated) purines or pyrimidines constitute the main components of DNA/RNA. Moreover, these heterocycles are common motifs in marketed drugs, natural products, and synthetic bioactive compounds (Fig. 1a)1,2,3,4. Therefore, their synthesis is widely recognized as a key objective for carbohydrate synthesis in laboratory and industrial settings.

Fig. 1: The importance of N-glycosides and the background of this research.
figure 1

a Representative examples N-glycosides as marketed drugs and bioactive compounds; b Selected examples of ionic glycosylation; c previous N-glycosylation via radical activation; d this work: dehydroxylative radical N-glycosylation.

Full size image

Despite the existence of other methods, including de novo approaches and the coupling reactions of glycosyl amines5,6,7,8,9, N-glycosylation offers a straightforward strategy to access N-glycosylated heterocycles due to the ubiquity of starting materials 10 N-glycosylation typically occurs through an ionic process, in which the glycosyl donor undergoes activation to generate a highly electrophilic intermediate (e.g., oxocarbenium) and then react with an N-heterocycle (Fig. 1b). Techniques following this activation mode typically necessitate the use of labile glycosyl donors and harsh reaction conditions, as exemplified by elevated temperatures in fusion reactions11, labile glycosyl halides and highly basic conditions in metal salt-based methods12, and acidic conditions in Vorbrüggen reactions, thus causing functionality compatibility problems (Fig. 1b)13. Recent advancements by Yu14,15, Hotha16, Chen17 and Xiao18 have facilitated N-glycosylation under significantly milder conditions, offering dependable tools for addressing diverse synthetic challenges (Fig. 1b)19,20,21,22. However, costly catalysts and/or requisite glycosyl donors are still required in these approaches. Therefore, it remains highly desirable to develop general N-glycosylation methods that use stable glycosyl donors, circumvent the use of expensive catalysts or promoters, and avoid harsh activation conditions.

The development of mechanistically distinct glycosylation methods may provide opportunities to address some of the aforementioned limitations. Considering the mild conditions, great functionality tolerance, and unique stereoselectivity, a glycosyl radical-based approach would be intriguing23,24,25. However, great challenges exist for realizing such an N-glycosylation, including common issues such as stereoselectivity and regioselectivity, as well as the tendency of glycosyl radicals to undergo homocoupling26,27, oxidation28 or reduction29. Consequently, while the synthesis of C-glycosides23,24,25 and S-glycosides30,31,32 has been well-established, research on corresponding N-glycosylation remains virtually untouched. Recently, Niu et al. demonstrated the possibility of radical N-glycosylation by utilizing allyl glycosyl sulfones as glycosyl donors28. However, their investigation was limited to only two substrates, and the formation of the C-N bond still followed an ionic SN2 process via glycosyl iodides (Fig. 1c)28.

Inspired by recent advances in copper-catalyzed C-N coupling reaction33,34,35,36,37,38,39 and dehydroxylative coupling reactions40, specifically employing metallaphotoredox catalysis41, we sought to develop a copper-catalyzed, dehydroxylative radical N-glycosylation as an alternative to conventional ionic approaches. In our working hypothesis (Fig. 1d), we envisioned that a glycosyl radical could be generated from a 1-hydroxycarbohydrate after proper activation40 under mildly photoredox conditions23,24,25,42. It can then capture the copper (II)-amido complex efficiently43. Subsequently, the resulting high-valent Cu(III) species can undergo facile reductive elimination to yield the desired N-glycosides44,45. We report herein the successful realization of this hypothesis (Fig. 1c). Merits of this work include i) mechanistically distinct, dehydroxylative radical N-glycosylation; ii) readily available and stable 1-hydroxycarbohydrates as the glycosyl donors and inexpensive copper catalyst; iii) highly diastereoselective and regioselective for a broad substrate scope.

Results and discussion

Though alkyl halides and carboxylic acids have been extensively studied in C-N coupling reactions33,34,35,36,37,38,39, the corresponding glycosyl counterparts are either unstable or difficult to prepare46. Therefore, our objective was to directly activate 1-hydroxycarbohydrates to form glycosyl radicals. Following a preliminary investigation of hemioxalate47, dihydropyridine (DHP) ester48,49 and xanthate salt50 activation techniques (see section 2.4 in Supplementary Information), we decided to use the N-heterocyclic carbene (NHC) activation method developed by MacMillan and co-workers (Supplementary Table 13)51,52,53. This method does not require isolation or workup to prepare the NHC-adduct, making it the preferred choice.

After a comprehensive screening process, we successfully identified the optimal reaction conditions, leading to a yield of 87% for 0.1 mmol and 74% for 1 mmol scale of the target product 2a (see Supplementary Table 1–14 and section 2.3 in Supplementary Information). The reaction protocol consisted of using [Cu(CH3CN)4]PF6 as the catalyst, 4,4’-di-tert-butyl-2,2’-bipyridine (dtbbpy) as the ligand, 2,4,5,6-Tetra(9H-carbazol-9-yl)isophthalonitrile (4CzIPN) as the photocatalyst, CsOAc as the base, and t-BuOOH as the oxidant in MeCN (entry 1, Table 1). Control experiments demonstrated that copper catalyst, photocatalyst, base, oxidant, light and ligand were all indispensable for the reaction (entries 2–7). Replacement of [Cu(CH3CN)4]PF6 with Cu(TMHD)2 yielded the product in 66% yield as a mixture of N1/N2 regioisomers (entry 8). Regioselectivity was not compromised when 1,10-Phenanthroline (Bphen) was used, but the yield decreased (entry 9). Other bases such as 2-tert-butyl-1,1,3,3-tetramethylguanidine (BTMG) displayed unsuitability for the reaction (entry 10). Though cumyl trimethylsilyl peroxide (cumOOTMS) and Ir-based photosensitizer proved to be suitable for the reaction, they are not selected due to their high cost (entries 11-12). The reaction efficiency significantly decreased in DMSO (entry 13). Moreover, the reaction exhibited tolerance to temperature variations (up to 50 °C) and moisture (entries14–16); 78% yield was obtained with the deliberate addition of 10 equivalents of water. The fact that the tolerance of some extent of water, a potential advantage relative to ionic N-glycosylation, is not clear at the moment54. Other glycosyl donors derived from 1a such as 1c-1e cannot afford any products under the optimized conditions (entries 17-19). Finally, common alcohols (1f-1g) could not react under the optimized conditions, suggesting the unique role of glycosyl radicals in this transformation. It is noteworthy that while 1-hydroxy carbohydrates have been utilized as glycosyl donors for O-55 and C-glycosylation51, we demonstrate the successful application of 1-hydroxy carbohydrates for N-glycosylation in this work.

Table 1 Optimization of the Reaction Conditionsa
Full size table

After determining the optimized conditions, we proceeded to investigate the range of N-heterocycles using furanose 1a (Fig. 2). To our delight, the reaction is successful with over 10 different N-heterocycles, including indazole (2a-2i), azaindazole (2j-2k), pyrazole (3a-3g), triazole (3 h), tetrazole (3i), carbazole (4a-4g), pyrrole (4h-4i), indole (5a-5j), azaindole (5k-5n), pyrimidine (6a-6c), and purine (6d-6g), cyclic imide (7a), β-lactam (7b), isoquininone (7c), and cyclic carbamate (7d). α-N-glycosides were consistently observed in all cases, showing 1,2-trans selectivity similar to other glycosyl radical-based glycosylations for furanose substrates56,57,58,59,60,61, likely due to steric effects. Different from indazoles, pyrazoles proved to be problematic substrates, resulting in the formation of 2:1 regioisomers (3a). This issue was found to be solvable through re-optimization using the copper(I) thiophene-2-carboxylate (CuTc) and 4,7-diphenyl-1,10-phenanthroline (Bpen) catalytic system (regioselectivity>20:1). Other heterocycles containing multiple N-atoms exhibited excellent regioselectivity (e.g. 3b-3e, 3 h, 3i, 6d-6g). Importantly, purines yielded only N9-glycosides without any N7-side products (6d-6g)7. The reaction conditions are mild and can accommodate various functionalities, including halo (2a, 2 f, 2 g, 2i, 4b, 4e-4f, 5d-5h, 5l-5n), nitrile (2e, 4c, 5c), ester (2d, 3d, 3 g, 3 h, 4i, 5a, 5b), keto (2c, 4 h), ether (2 h, 4d, 5i), trifluoromethyl (3e, 3 f, 5j), and common protecting groups [N-Bz (6a-6c), N-Boc (6e, 6 f), N-Bn (6d), O-carbamate (6 g)]. It is worth noting that the bromo-group remains intact (2i, 4b, 4 f, 5d, 5 h, 5i), providing a reactive site for subsequent derivatization through other metal-catalyzed cross-coupling reactions. Our protocol provides a convenient alternative for accessing certain significant N-glycosides. For instance, glycosyl carbazoles usually require the utilization of 1,2-anhydro sugars as glycosyl donors followed by the construction of the carbazole or a de novo synthesis of the sugar ring62,63. Furthermore, to our knowledge, there is no prior report on the direct N-glycosylation of β-lactam, a privileged motif found in antibiotics. Finally, it should be noted that our method has certain limitations; in particular, (benzo)imidazole cannot be involved, and indoles with electron-rich groups are not able to efficiently participate in this reaction, possibly due to the decreased acidity of N-H group.

Fig. 2: The substrate scope with respect to N-heterocycles.
figure 2

Conditions: a TMG as base. b CuTc (10 mol%), Bphen (10 mol%). c Cu(TMHD)2 (10 mol%). no ligand. d DMSO as solvent. N.D. Not detected, Ac Acetyl, Bz Benzoyl, Bn benzyl, Boc t-Butyloxy carbonyl.

Full size image

We moved on to explore the scope of sugar substrates with 1b unchanged (Fig. 3). Due to the tendency of NHC-adducts derived from sugars other than 1a to decompose at room temperature, we conducted the reaction at 0 °C to ensure its effectiveness. Furanose derivatives with a range of protecting groups consistently produced 1,2-trans products (8a-8l), except for 2-deoxyribose lacking a substituent at C2 (8k). Intriguingly, the presence of a small fluorine atom can restore the trans-selectivity, possibly due to the fluorine conformational effect (8j)64. The electron-rich protecting group is beneficial to the reaction, and perbenzoyl protected 8 g was obtained in a much lower yield than 8 h. This argument is also supported by the fact that acyl-protected pyranoses do not participate in the desired transformation, while only those with ether-type (e.g. OMe) protecting groups do (8m-8p). Among the pyranose series, substrates derived from, mannose (8 m), galactose (8n), and rhamnose (8o) predominantly yielded the α-isomer. This can be explained by stereoelectronic effect65,66,67, which refers to the hyperconjugative interaction between the lone pair of the endocyclic oxygen and the low-lying C−Cu antibonding orbitals. This interaction leads to the preference of the copper species at the axial position, resulting in high α-stereoselectivity. The glucose-derived substrate was obtained as a mixture of anomers (8p), perhaps due to the glucosyl radical adopting a B25 conformation, in which the steric hindrance of C2-OMe reduces α-selectivity65,66. Nevertheless, it is important to recall that highly β-stereoselective C-glycosylation with glucosyl substrates can be achieved under Ni-catalyzed conditions68,69.

Fig. 3: The substrate scope with respect to sugar and the late-stage modification.
figure 3

Conditions: a TMG as base. b CuTc (10 mol%), Bphen (10 mol%). c Cu(TMHD)2 (10 mol%). no ligand. d DMSO as solvent. e at 0 °C. Bz Benzoyl, Bn benzyl, Tr Trityl, MOM Methoxymethyl.

Full size image

To demonstrate the applicability of this N-glycosylation protocol, we sought to functionalize a wide variety of structurally complex molecules in a late-stage fashion. Marketed anticancer drugs, including Olaparib (9a), Tegafur (9b), and Axitinib (9d), underwent successful conversion to the corresponding N-glycosides with yields ranging from 47% to 58% as the exclusive α-isomer. Metaxalone, a muscle relaxant (9c) and Losartan (9e), an anti-hypertension drug, are also viable substrates. Interestingly, the latter compound demonstrated that the presence of a free alcohol group does not affect the reaction efficiency. Finally, the natural product rutecarpine, derived from the traditional Chinese medicine fructus evodiae, underwent efficient N-glycosylation, resulting in the formation of (9f) with a 53% yield.

To understand the reaction pathway, we have conducted a series of mechanistic studies (see section 2.5 in Supplementary Information). The reaction was completely shut down by the addition of TEMPO, thus suggesting a possible radical pathway (Fig. 4a). When 10 with an allyl group at O-2 position was used, diastereomers of 11 were obtained in 38% yield, which presumably arises from 5-exo-trig cyclization of the radical intermediate (Fig. 4b). This radical clock reaction clearly proved the involvement of glycosyl radical intermediates. The reaction cannot occur without photosentizer or light (Entries 3, 5, Table 1). Furthermore, the Stern–Volmer experiments showed that NHC-alcohol adduct quenched the excited state photocatalyst (PC*) significantly faster than other reagents such as an indazole (Fig. 4c). These experiments, along with the control experiments listed in the reaction optimization section (Table 1), suggested glycosyl radicals are likely generated by photoactivation.

Fig. 4: Mechanistic Studies.
figure 4

a The radical quenching experiment. b The radical clock experiment. c Stern-Volmer quenching experiments. d UV-vis absorption spectroscopy.

Full size image

We next conducted UV–vis absorption spectroscopy and cyclic voltammetry experiments (see section 2.5.3-2.5.4 in Supplementary Information) to get some insight on the oxidation of Cu (I) to Cu (II) in the copper catalytic cycle (Fig. 4d). According to the UV/Vis studies, reaction of LnCu(I) with t-BuOOH generated a new copper species (i→ii), which has a nearly identical profile as LnCu(II) (iii). The oxidation of LnCu(I)-amido to LnCu(II)-amido (iv→v) is also feasible, probably easier than the oxidation of LnCu(I) to LnCu(II), as shown by the CV experiment (Supplementary Fig. 7).

On the basis of the conducted experiments, we proposed a plausible mechanism as illustrated in Fig. 5. The reaction is initiated by the Single Electron Transfer (SET) oxidation of the NHC adduct I with 4CzIPN51, resulting in the formation of glycosyl radical II. In the copper catalytic cycle, the ligand change reaction of the copper catalyst with an N-heterocycle (e.g., 1b) generates copper (I)-amido complex III. This species is then oxidized by the oxidant (tBuOOH), leading to the formation of copper (II)-amido species IV. The glycosyl radical is then captured by IV, yielding copper (III) complex V43, which gives the desired product through reductive elimination44,45, thus regenerating the copper catalyst and closing the copper catalytic cycle. The photoredox cycle is closed by the oxidation of the reduced form of the photocacalyst (PC-) with the oxidant (tBuOOH), followed by sensitization.

Fig. 5
figure 5

Proposed Mechanism for the dehydroxylative radical N-glycosylation.

Full size image

In summary, we have developed a dehydroxylative radical method for synthesizing N-glycosides by leveraging copper metallaphotoredox catalysis, in which stable and readily available 1-hydroxy carbohydrates are directly activated for N-glycosylation. Complementing with the well-established ionic approaches, our method employs inexpensive photo- and copper- catalysts and can tolerate some extent of water. The reaction exhibits a broad substrate scope, encompassing 76 examples, and demonstrates high stereoselectivity, favoring 1,2-trans selectivity for furanoses and α-selectivity for pyranoses. It also exhibits high site-selectivity for substrates containing multiple N-atoms. The synthetic utility was showcased through the late-stage functionalization of bioactive compounds and pharmaceuticals like Olaparib, Axitinib, and Metaxalone. The presence of glycosyl radicals was confirmed through radical suppressing reactions and a radical clock reaction. Additionally, the importance of copper metallaphotoredox catalysis was demonstrated through control experiments and various spectroscopic studies, such as UV-vis experiments. Though limitations, such as unsuitability for certain electron-rich hetercoycles and electron-deficient sugar substrates, still exist, we believe this work will stimulate more research in the radical N-glycosylation for the preparation of valuable N-glycosides that are difficult-to-made in future.

Methods

For 1H, 13C, and 19F nuclear magnetic resonance (NMR) spectra of compounds in this manuscript and details of the synthetic procedures as well as more reaction condition screening, see Supplementary Information.

General procedure for 2a-2k, 3d, 3f, 3g, 8l. For more substrate procedures see Supplementary Information

An oven-dried 10 mL Schlenk tube was charged with 1-hydroxylmannose 1a (93.6 mg, 0.36 mmol, 1.8 equiv), NHC (142.4 mg, 0.36 mmol, 1.8 equiv) and a magnetic stir bar. After the Schlenk tube was vacuumed and refilled with nitrogen gas three times, dry toluene (2.0 mL) was added and the reaction was stirred at r.t. for 5 min. Then, pyridine (29.1 µL, 0.36 mmol, 1.8 equiv) was added dropwise at room temperature. The resulting solution was stirred at r.t. for 10 min. A white solid precipitated out during this time. Another 10 mL Schlenk tube was charged with 1,2,3,5-Tetrakis(carbazol-9-yl)−4,6-dicyanobenzene (4CzIPN, 5 mg, 0.01 mmol, 0.05 equiv), Cu(MeCN)4PF6 (7.4 mg, 0.02 mmol, 0.1 equiv), dtbbpy (4,4-di-tert-butyl bipyridine, 8.0 mg, 0.03 mmol, 0.15 equiv), CsOAc (38.2 mg, 0.2 mmol, 1.0 equiv), N-heterocycle (0.2 mmol, 1.0 equiv) and a magnetic stir bar. This Schlenk tube was vacuumed and refilled with nitrogen gas three times. Dry acetonitrile (2.0 mL) was added to this Schlenk tube under an atmosphere of nitrogen and stirred at room temperature. The toluene suspension was transferred to a 5 mL syringe under an atmosphere of nitrogen. Then a syringe filter and new needle were installed on the syringe, and the toluene solution was injected through the syringe filter into the MeCN solution. t-BuOOH (80 μL, 5-6 M in decane, 2.0 equiv) was added, before subjecting the reaction mixture to irradiation by 420 nm blue LEDs at room temperature for a duration of 4 hours. The organic layers were evaporated and then purified by flash column chromatography on silica gel.