Introduction

Nucleosides play vital roles in a myriad of physiological processes, including enzyme regulation and metabolism, cell signaling, as well as DNA/RNA synthesis. It is therefore not surprising that nucleoside derivatives mimicking their endogenous counterparts have been widely used for disease treatment through various functional mechanisms, including DNA/RNA synthesis inhibition1 and gene interference (gene silence and RNA interference)2,3. For instance, nucleoside derivatives account for almost half of the available antiviral medicines4,5, while 15 nucleosides have been approved by the FDA for clinical cancer chemotherapy5,6,7. Nucleoside derivatives have also contributed greatly to curb the epidemic of COVID-19, as many FDA-approved and clinically tried anti-COVID-19 drugs are nucleosides (remdesivir and molnupiravir)8,9. Meanwhile, nucleosides creating artificial extra base pairs beyond A-T and G-C occupy the central position in explorations of the genetic alphabet expansion, opening the door to establish novel biological systems and next-generation biotechnologies10. Moreover, as secondary metabolites, complex nucleosides with evident structural variations from endogenous nucleosides have been demonstrated to exhibit significant bioactivities, especially antibacterial and antifungal activities11. As such, the multifunctionality of nucleoside derivatives warrants the intensive efforts devoted to the development of efficient protocols for nucleoside preparation, and synthesis of noncanonical nucleoside derivatives has been deemed as an emerging area of high potential impact recently12. Continuous endeavors have culminated in the establishment of three synthetic strategies13,14: N-glycosylation, sugar-ring construction, as well as enzymatic transglycosylation strategies.

As the canonical and mainstay strategy and represented by the Vorbruggen protocol15, the anomeric cation-involved N-glycosylation strategy has witnessed tremendous progress, as evidenced by the applied glycosyl donors evolving from conventional glycosyl acetates15, thioglycosides11, and trichloroacetimidates (TCI)16 to the cutting-edge glycosyl o-alkynylbenzoates (ABzs)17, 3,5-dimethyl-4-(2′-phenylethnylphenyl)phenyl (EPP) glycosides18, glycosyl o-(1-phenylvinyl)benzoate (PVB)19, as well as glycosyl carbonate donors (Fig. 1)20. Additionally, with the renaissance of the radical-based glycosylation21, N-glycosylations entailing glycosyl anomeric radical species also appeared sporadically22,23. Although big stride has been made, the drawbacks, such as a stoichiometric promoter system, highly difficult-to-operate nucleobase preactivation (silylation), limited substrate scope, sluggish reaction speed, and unsatisfactory synthetic efficiency, need to be appropriately tackled to improve the synthetic efficiency. The sugar-ring construction strategy relies on intramolecular glycosylation24, intramolecular substitution25, RCM reaction26, or iodocyclization to construct the sugar ring27,28. However, the challenging C-N bond formation cannot be eliminated; instead, it is moved before sugar-ring fabrication. Therefore, the drawbacks associated with the N-glycosylation strategy still remain in the sugar-ring construction strategy. In addition, the synthesis of ring-closure precursor and post-ring-closure modifications is generally a multistep process, further compromising the synthetic efficiency. As a useful complement to the chemical strategies, the enzymatic transformation strategy has made substantial progress in the past decades. Nevertheless, the high reliance on two kinds of enzymes, the nucleoside 2′-deoxyribosyltransferases (NDTs) and nucleoside phosphorylases, some of which are difficult to access, seriously restricts the synthesis scale and product diversity14. Under such a context, novel synthetic protocols exquisitely solving the drawbacks mentioned above are highly desirable29.

Fig. 1: Existing strategies for nucleoside synthesis and the BEPT protocol.
figure 1

A N-glycosylation strategy via either ionic or radical mechanisms (LG leaving group, o-ABz ortho-alkynylbenzoyl, EPP 3,5-dimethyl-4-(2′-phenylethnylhenyl)phenyl, PVB ortho-(1-phenylvinyl)benzyl, EGC ethynylcyclohexyl glycosyl carbonate). B Sugar-ring construction strategy (LG leaving group, NB nucleobase). C Enzymatic transglycosylation strategy (NB nucleobase). D The BEPT protocol (BEP o-tert-butylethynylphenyl).

In this work, inspired by the successful application of o-methoxycarbonylethynylphenyl thioglycosides (MCEPTs) in N,O-glycosides synthesis30 and with o-(tert-butylethynyl)phenyl thioglycosides (BEPTs) as donors, an alternative BEPT protocol for efficient nucleosides synthesis is established. The BEPT protocol is distinguished from existing methods by enhanced efficiency, expanded substrate scope, dramatically simplified experimental operation, lowered catalyst loading, and accelerated reaction speed. Combined with the saccharide C-O bond editing techniques31, stereoselective and concise synthesis of the challenging 2′-deoxy nucleoside derivatives was also achieved via the BEPTs protocol32. Through the substrate scope investigations, the beneficial solvent effects of hexafluoroisopropanol (HFIP)33 in nucleoside synthesis, including dramatically improving the reaction speed in the protecting group (PG)/solvent procedures and accommodating free uracil derivatives as viable acceptors in the solvent procedure, are disclosed. The protocol is amenable to scale-up synthesis of nucleosides, and is successfully applied in efficient and convergent/flexible syntheses of the challenging but biologically relevant nucleosides angustmycin A, dJ, and dJ 3′-deoxy analog. Moreover, the underlying reaction mechanisms are also investigated, providing deep insights into the BEPT protocol.

Results and discussion

The investigation commenced with glycosylation of 2a with MCEPT donor 1a’ under the catalysis of Ph3PAuNTf2 (Fig. 2A). Unexpectedly, no desired glycosylation product 3a was detected in the presence of 0.2 equiv of Au(I) catalyst in 4 h, and all donors were recovered. The failure of donor 1a’ in glycosylation of adenine derivative 2a was attributed to the electron-withdrawing effect of the methoxycarbonyl group in MCEPT donor, which decreased the electron density of C≡C triple bond thus weakening its coordination with Au(I) catalyst; meanwhile, as a competitive interaction the coordination between Au(I) catalyst and basic nucleobase acceptor predominated thereby halting the glycosylation reaction. Nevertheless, the methoxycarbonyl group has been proven essential in validating o-alkynylphenyl thioglycoside-type donors as it provides critical stabilization to the pivotal catalyst species—the benzothiophen-3-yl gold(I) complex—to suppress orthoester by-product formation30. This dilemma raised the question of whether a steric alkyl substituent, rather than an electron-withdrawing group, could simultaneously stabilize the benzothiophen-3-yl gold(I) intermediate and enhance the affinity between C≡C and Au(I) catalyst so as to validate a novel type of glycosylation donor with mitigated undesired interference of nucleobase acceptors. To test this hypothesis, the methoxycarbonyl group of 1a’ was replaced with a bulky but electron-donating tert-butyl (tBu) group, resulting in BEPT donor 1a (see Supplementary Information), which was then subjected to condense with 2a under otherwise identical conditions as used for 1a’. Pleasantly, the desired glycosylation products 3a and 3a’ (with one tert-butoxycarbonyl (Boc) group cleaved) were obtained in a 94% combined yield (3aa: 52%, 3aa’: 42%) along with the departed leaving group I (96%). The high glycosylation efficiency of BEPT donor 1a with 2a verified the conjectured favorable effects of the tBu substituent in BEPT donors, that is, the stabilizing effect on pivotal Au(I) catalyst intermediate and the activating effect on leaving group, with the former fully corroborated by the isolation and thorough characterization of the tert-butyl-substituted benzothiophen-3-yl gold(I) intermediate II (CCDC: 2414077, Fig. 2B). To shed further light on the reaction mechanism, a series of NMR experiments were conducted. First, the interaction between nucleobase 2a’ and Au(I) was investigated (Fig. 2C). Upon addition of 0.2 equiv of Au(I) catalyst to the solution of 2a’ in deuterated chloroform in an NMR tube, an evident downfield shift of H-8 from 8.33 to 8.51 ppm was detected, demonstrating the complexation between Au(I) and 2a’. The H-8 signal shifted gradually to the downfield with the amount of Au(I) catalyst increased and finally reached 9.10 ppm at 1.0 equivalent of catalyst. Increasing the catalyst amounts from 1.0 to 1.5 equivalents did not cause the signal of H-8 to shift downfield further, indicating the stoichiometry of catalyst and nucleobase in Au(I)-nucleobase complex was 1: 1. According to the donor design considerations and results of model reaction, the BEPT donor should compete with nucleobase to coordinate with Au(I) catalyst, leading to donor activation and nucleobase release simultaneously and ultimately product formation. To confirm this, NMR titration of 2a’ and Au(I) catalyst (0.5 eq) with dramatically simplified BEPT donor III (see Supplementary Information) was performed. Indeed, with the addition of III, the H-8 signal shifted upfield from 8.77 to 8.52 ppm (20 equivalent of III), demonstrating the gradual release of nucleobase 2a’ under the effect of III (Fig. 2D). For direct comparison, solutions of 2a in the presence of 0.2 equivalents of Au(I) catalyst were added 2.0 equivalents of III and simplified MCEPT donor IV (see Supplementary Information), respectively. Instead of upfield shift, a slight downfield shift of H-8 of 2a was observed in the presence of IV indicating the negligible interaction between IV and Au(I) catalyst in the presence of nucleobase (Fig. 2E). From Fig. 2C–E, an interesting phenomenon that even under the effect of 0.2 or 0.5 equiv of Au(I) complex only one species of nucleobase instead of two was detected from the NMR spectra. Given the 1:1 stoichiometry of catalyst and nucleobase in Au(I)-nucleobase complex determined in Fig. 2C, this observation is really intriguing. The rapid reversible coordination/dissociation between the Au(I) catalyst and nucleobase might be invoked to interpret this abnormal observation, which made the free and coordinated 2a’/2a indistinguishable in the NMR time scale, thus presenting average spectra in all cases. Along this line, a conclusion that the deactivation of Au(I) catalyst by nucleobase acceptor is not the irreversible poisoning but the rapid reversible coordination/dissociation, which considerably decreases the effective concentration of catalyst, could be deduced. Indeed, under 0.6 equiv of Au(I), glycosylation of 2a could also be achieved by the MCEPT donor 1b’ to deliver the nucleosides in a 95% combined yield (Fig. 2A). DFT calculations were subsequently carried out to decipher the coordination pattern of Au(I) catalyst and 2a and to determine the affinity difference between Au(I) catalyst and 2a/III/IV, respectively (Fig. 2F). Although N3, N7, and N9 (corresponding to N1-3 in the DFT study) all could coordinate with Au(I) catalyst, the N7-liganded complex P12 exhibited the highest stability with the ΔG being −16.16 kcal/mol. Compound III could also coordinate with Au(I) catalyst efficiently as the coordination ΔG reached as low as −11.16 kcal/mol (P13), higher than N7-ligated complex but remarkably lower than N3- and N9-ligated complexes (P12 and P12′′). In contrast, the coordination of IV with Au(I) catalyst was much less thermodynamically favorable than III (P14, ΔG = −1.86 kcal/mol), disclosing the underlying factor responsible for the invalidity of MCEPT donor in nucleobase glycosylation at low catalyst loading.

Fig. 2: Mechanism elucidation of the BEPT protocol in nucleoside synthesis.
figure 2

A Glycosylation of 2a with MCEPT and BEPT donors (MCEPT o-(methoxycarbonylethynylphenyl)thio). B Isolation of the key catalyst intermediate II. C The chemical shift migration of H-8 of 2a’ with the increment of Au(I) catalyst amounts. D Titration of 2a’ and Au(I) catalyst (0.5 equiv) with III. E Comparison of the chemical shifts of H-8 in 2a in the presence of the same amount of III and IV (2.0 equiv). F DFT calculations (The calculations were conducted using Gaussian 09D).

Combined with the control reactions, the NMR experiments, and the DFT calculations illuminated the mechanism of N-glycosylation with BEPT donor, and meanwhile, spurred the following in-depth explorations to establish the BEPT-based protocol for efficient nucleoside preparation.

Encouraged by the promising results of the model reaction, the substrate scope of the BEPT-based nucleoside synthesis protocol was checked. As direct glycosylation of purine nucleobases avoiding prior activation/silylation could be realized by using Boc-protected purine derivatives as acceptors (PG procedure), the substrate scope exploration began with purine nucleobases (Fig. 3). Adenine and guanine derivatives 2a-f, featuring di-Boc-protected exocyclic amino groups, were chosen as representative purine acceptors to condense directly with BEPT donors derived from a variety of sugars, including D-glucose (1a-c), D-galactose (1d), D-mannose (1e), L-rhamnose (1f), L-arabinose (1g), D-xylose (1h), D-ribose(f) (1i), L-arabinose(f) (1j), D-fructose(f) (1k), and disaccharide lactose (1l), under the catalysis of Au(I) complex without need for purine base activation in advance. The di-Boc-protected nucleosides (3aa-3lf) and their mono-Boc-protected analogs (3aa’-3lf’) were obtained in good to excellent combined yields (59%–99% yields). It should be pointed out that the glycosylation systems were clean, and the di-Boc and mono-Boc nucleosides could be easily separated by silica gel column chromatography due to the disparate polarity. Interestingly, a conformation flip from 4C1 to 1C4 was observed for the mannosyl moieties in 3ec/3ec’ and 3ed/3ed’. Similar phenomenon was also detected in rhamnosyl nucleosides 3fb’, 3fc/3fc’, and 3fd, entailing 1C4 to 4C1 conformation flip (J1′,2′ > 5.0 Hz). Ketosyl BEPT glycoside 1k served also as a competent donor to glycosylate di-Boc-protected purines, providing 3ka/3ka’ and 3kd/3kd’ in 96% and 95% yield, respectively. Noticeably, lactosyl BEPT glycoside 1l also proved to viable donor, furnishing disaccharide nucleosides 3lf/3lf’ in good yield (79%). Apart from the couplings of 1d/2a and 1l/2f, which required 24 h to reach completion, all other tested condensations finished within 4 h, faster than related Au(I)-catalyzed nucleoside synthesis protocols17,20. Moreover, some glycosylation reactions, such as those involving the donor/acceptor pairs of 1b/2b, 1b/2c, 1b/2d, 1c/2d, 1d/2b, and 1i/2d, were completed in just 1 h, highlighting the high reaction rate of the BEPT protocol. The isolation of mono-Boc-protected nucleosides implies the high propensity of the di-Boc-protected amino group to undergo one Boc cleavage under mild conditions, which could be exploited to simplify product purification by post-glycosylational one-pot conversion of di-Boc- to mono-Boc-protected products. Specifically, upon completion of glycosylation, direct addition of silica gel to the reaction mixture of 1j and 2a and stirring delivered mono-Boc 3ja’ (74%) as the only product, as with the case for the coupling of 1j and 2d (3jd’: 93%). Remarkably, synthesis of purine-type nucleosides through the BEPT protocol could also be achieved under dramatically reduced-catalyst loading (0.02 or 0.04 eq) without an evident drop in yields (above 84%), albeit in extended reaction time (8–48 h). Fortunately, the prolonged reaction time unified the glycosylation products rendering mono-Boc nucleosides (3bb’, 3cc’, 3fb’, 3ib’, 3id’, 3jb’, 3jd’, and 3lb’) as the only products, except for 3fd’ and 3ld which were accompanied by NH2-free nucleoside 3fd” (92% combined yield) and 3ld’ (84% combined yield), respectively. Further investigations revealed that the PG procedure of the BEPT protocol enjoyed a profound solvent effect; with HFIP/dichloromethane (DCM) as solvent34, the rate of the couplings was dramatically enhanced making the couplings conducted under reduced-catalyst loading (0.02–0.04 equiv) to complete in 1 h to deliver the desired products in good to excellent yields (above 59% for 3bb/3bb’, 3ec/3ec’, 3fd/3fd’, 3hd/3hd’, 3ib/3ib’, 3kc/3kc’, and 3lf/3lf’). The favorable solvent effect of HFIP could be attributed to its acidity34, which facilitates the protodeauration, thus improving the catalyst turnover17. All the tested donors were demonstrated to be exceptionally stable, remaining virtually intact even after a prolonged time of storage at room temperature without special protection from air and moisture. The N-9 regioselectivity was determined by the 13C chemical shift of C-5 in nucleobase (bigger than 120 ppm)35, which was further confirmed by HMBC (see Supplementary Information).

Fig. 3: Direct synthesis of purine-type N-glycosides through the PG procedure of the BEPT protocol.
figure 3

Boc-protected purine derivatives were selected as acceptors. Boc butoxylcarbonyl.

Direct glycosylation of pyrimidines, omitting the prior activation/silylation, has not been documented. Akin to purine derivatives, the PG procedure was attempted to realize the direct synthesis of pyrimidine-type nucleosides. p-Methoxybenzyl (PMB) group was selected as a PG for the endocyclic N3 of uracil derivatives, while Boc was chosen as PG for the exocyclic NH2 of cytosine. Thus, N3 PMB-protected uracil (4a), thymine (4b), 5-fluorouracil (4c), and di-Boc-protected cytosine (4d) were selected to react with various BEPT donors derived from both hexoses and pentoses (in pyranosyl or furanosyl form, Fig. 4A). With 4a and 4d as acceptors, all tested donors presented N-glycosides 5ba-5id with yields exceeding 73% in 1 h under the catalysis of 0.2 equivalents of Au(I) complex, achieving direct, swift, and highly efficient synthesis of pyrimidine-type nucleosides. Sugar conformation flips were also noticed for N-glycosides 5ea, 5ed, 5fa, and 5fd, derived from D-mannosyl (1e) and L-rhamnoseyl (1f) BEPT donors, as confirmed by 1H NMR spectra (J1′,2′ > 5.0 Hz) and single crystal X-ray diffraction analysis (CCDC: 2401831 for 5fa). In addition, mono-Boc products were not detected when 4d was used as the acceptor. In sharp contrast, when 4b,c, the 5-substituted analogs of 4a, were used as acceptors, the O-glycosides 6bb-6ic were isolated (above 75% yields), as determined by the 13C NMR (above 90 ppm for the anomeric carbons) and HMBC (correlations between anomeric H and C2 of nucleobase, yet non-correlations between anomeric H and C6 of nucleobase). Moreover, although equipped with C2-O-appended benzoyl groups amenable to exert neighboring group-participation effect, perbenzoylated glucosyl and ribofuranosyl BEPT donors 1b and 1i afforded 6bb/6ic as mixtures of α/β isomers, while 6bc α-exclusively. Prolonging the reaction time of 1b and 4b to 12 h under otherwise identical conditions allowed for the generating α-6bb exclusively in high yield (90%), while 0.02 equivalents of catalyst afforded β-6bb only (see Supplementary Information). These results suggest an in situ epimerization of β-isomer to α-isomer during glycosylation reaction, presumably through the sugar-ring opening/reclosing (at C1-O5 bond) mechanism36. The generation of O-glycosides instead of N-glycosides was ascribed to the steric hindrance caused by N3- and C5-located substituents, which made N1 difficult to access. Thus, even under the conventional silylation/glycosylation conditions, the condensation between 1b and 4b still gave O-glycoside as the major product (see Supplementary Information). However, an exception was also observed when 4b was condensed with ribofuranosyl BEPT donor 1i, whence N-glycoside 5ib was secured (95%), presumably due to the synergy of the high glycosylation potential of 1i and ameliorated reactivity of 4b with an electron-donating methyl substituent. The PMBs in N-glycosides could be smoothly cleaved under the effect of ceric ammonium nitrate, as exemplified by 5ca, 5fa, and 5ia, which delivered 5ca’, 5fa’, and 5ia’ smoothly. On the contrary, attempts to remove the PMBs in O-glycosides met with failure, and glycosidic linkage rupture occurred prior to PMB cleavage. Efforts to convert O-glycosides to the corresponding N-glycosides under various conditions also proved futile, although it is well known that the O-glycosides can act as intermediates in nucleoside synthesis. Synthesis of pyrimidine-type nucleosides through the PG strategy of the BEPT protocol could also be conducted under reduced-catalyst loading (0.02 eq) to yield the desired nucleosides in good to excellent yields in 8 h (over 74% yield for 5ba, 5cd, 5dd, 5fa, 5hd, and 5ia), which could be shortened to 1 h in DCM/HFIP when 4d was used as acceptor (5cd, 85%).

Fig. 4: Direct synthesis of pyrimidine-type glycosides through the BEPT protocol.
figure 4

A Direct glycosylation of pyrimidine bases through the protecting group (PG) procedure of the BEPT protocol (PMB-protected pyrimidine bases were selected as acceptors. PMB p-methoxybenzyl. B Direct glycosylation of pyrimidine bases through the solvent procedure of the BEPT protocol (A mixture of DCM and HFIP (v/v = 4:1) was used as reaction media. DCM dichloromethane, HFIP hexafluoroisopropanol.

In continuous pursuit of direct and efficient synthesis of 5-substituted uracil nucleosides, a solvent procedure with HFIP as co-solvent was explored subsequently (Fig. 4B). The choice of HFIP is not only because of its verified beneficial effects on purine-type nucleoside synthesis but also because its well-known talent in H-bond formation, which at least could tackle the solubility problem of pyrimidine bases34. Indeed, free thymine was completely dissolved in DCM/HFIP leading to a homogeneous solution, while forming a suspension in pure DCM (Fig. 4B). Pleasantly, using DCM/HFIP as reaction media, glycosylation of 4e with 1b proceeded smoothly under the catalysis of Au(I) complex at room temperature to yield 7be in 80% yield; however, unexpectedly, the coupling reached completion in only 10 min. The procedure enjoyed broad substrate scope in terms of both glycosyl BEPT donors and 5-substituted uracil acceptors, yielding nucleosides 7bf-7jf in 71%–92% yields in 10–20 min. In fact, free uracil 4i also proved a viable acceptor, providing the desired nucleosides 7ci and 7ii in good yield. Subsequently, the reduced-catalyst-loading version of the solvent procedure was briefly studied, and high efficiency was maintained (7ce and 7ie). Interestingly, although free 4e-i were selected as glycosylation acceptors, no N3-regiomers were detected37. Directly using unprotected nucleobases as glycosylation acceptors, nucleoside synthesis through the solvent procedure eliminates not only silylation/activation processes but also PG manipulations, further streamlining the chemical synthesis of nucleosides. Again, conformational switch of D-mannosyl as well as L-rhamnosyl residues was noticed in the pyrimidine series nucleosides. Besides, BEPT donors were also integrable to the one-pot silylation/glycosylation procedure to give 5-substituted uracil-type nucleosides 7be-7if efficiently (above 86% yields), highlighting the versatility of the BEPT glycosylation method (see Supplementary Information).

The unexpected beneficial effects of HFIP in the solvent procedure were tentatively ascribed to its strong H-bond donation ability34, triggering the formation of favorable intermolecular H-bond systems between uracil derivatives and HFIP. This assumption was supported by DFT calculations (Fig. 5). In the presence of HFIP, an 8-membered intermolecular H-bond system was formed between thymine and HFIPs, leading to AG = −1.4 kcal/mol). Upon addition of the catalyst, the Au(I) complex promoted the formation of another 8-membered intermolecular H-bond system, resulting in B, which was highly thermodynamically favored (ΔG = −24.8 kcal/mol). The DFT calculations were further corroborated by the following 2D NMR experiments. In the absence of Au(I)-catalyst, 1-NH/HFIP OHs, 3-NH/HFIP OHs, and 3-NH/HFIP C-H correlations were detected in the NOESY spectrum, demonstrating the existence of A, which was further supported by the evident correlation between 1-NH and 3-NH enabled by the flexibility of 1-NH (see Supplementary Information). Under the effect of Au(I) catalyst, the emergence of the correlation between 1-NH and HFIP C-H and the disappearance of the correlation between 1- and 3-NHs were detected. In addition, the correlation between phenyl ring of the catalyst ligand and thymine methyl group in the NOESY spectrum revealed the coordination of pyrimidine base to the Au(I) catalyst through the C4-carbonyl oxygen. All of these spectroscopic proofs pointed to the formation of B. Effected by the two flanked carbonyl groups, the acidity of 3-NH should be much higher than that of 1-NH, which would inevitably make the 3-NH-involved cyclic H-bond system much stronger than the 1-NH-engaged cyclic H-bond system, thereby assuring the high N1 regioselectivity of the glycosylation reactions. Moreover, in the 1-NH-involved cyclic H-bond system, the 1-NH acted as a H-bond donor to form the favorable H-bond with appropriate strength, thus resulting in improved reactivity of 1-NH in nucleoside synthesis38.

Fig. 5
figure 5

DFT calculations (The calculations were performed with B3LYP/def2-SVP//M06-2X/def2-TZVP).

Despite continuous efforts, the stereoselective synthesis of 2′-deoxy nucleosides remains a prominent challenge in carbohydrate chemistry39. The advent of the BEPT protocol provided a solution to this chronic problem upon cooperation with the photocatalyzed sugar C–O bond editing method31, with 1,4-dihydroisonicotinoyl (DHIN) group as both the stereo-controlling and the deoxygenation-initiating group (Fig. 6). To our pleasure, under the standard conditions, donors 1m,n bearing DHIN group at C2-OH reacted smoothly with purine/pyrimidine acceptors 2,4 under the PG procedure to produce 3ma-3nd β-stereoselectively in good yields (70–85% yield). The reaction time could be further shortened to 30 min by applying DCF/HFIP as reaction media even in the presence of 0.02 equiv of catalyst (84% for 3md and mono-Boc-protected 3md). Moreover, the subsequent deoxygenation also proceeded uneventfully to give 2′-deoxy-nucleosides 8ma-mc (69–71% yield) under the combined effect of 1,2,3,5-tetrakis(carbazol-9-yl)-4,6-dicyanobenzene (4CzIPN), adamantanethiol, and light (467 nm). Further investigations turned out that the deoxygenation was highly dependent on the nucleobase subunits, with nucleoside 5md derived from 4d providing the desired 2′-deoxynucleoside 9md (33%) accompanied by isonicotinoyl (IN) group-containing by-product 9md’ which was photo-catalytically inert (see Supplementary Information)40, while nucleosides 3md/3nd derived from acceptor 2d affording IN-containing by-products 8md’/8nd’ completely (70% and 72% yield). Surprisingly, DHIN-containing donors 1m,n could not condense efficiently with free uracil acceptors in HFIP/DCM. As a detour to access the corresponding 2′-deoxy nucleosides, the BEPT-involved one-pot silylation/glycosylation procedure was adopted, and the obtained 5-substituted uridines were successfully converted to 2-deoxy nucleosides under the photocatalytic conditions (62–85% yield) (see Supplementary Information).

Fig. 6
figure 6

Synthesis of 2′-deoxynucleoside derivatives by the combination of BEPT protocol and photocatalyzed C–O bond editing techniques (1,4-dihydroisonicotinoyl (DHIN) was used as a stereo-controlling and deoxygenation-initiating group, and the direct saturation of the DHIN group to isonicotinoyl (IN) group was found to be the major side reaction).

For direct comparison of the BEPT protocol with existing catalytic protocols in nucleoside synthesis (the ABz and EGC protocols), syntheses of 3ba/3ba’, 3ia/3ia’, and 3ja/3ja’ through the BEPT protocol were repeated under the catalysis of 0.1 equivalents of Au(I) catalyst (Fig. 7A). For 3ba/3ba’, the BEPT protocol afforded 81% yield of products much higher than that of the ABz protocol (3ba, 48% yield)17, as with the case for the synthesis of 3ia/3ia’, wherein the BEPT protocol provided 81% yield of 3ia/3ia’ higher than those of the ABz (3ia, 77%)17 and EGC protocols (3ia, 60%)20. At the very least, comparable yields of 3ja/3ja’ were registered between the BEPT and ABz protocols (82% vs 80%). Even under the decreased catalyst loading (0.04 eq), either comparable or higher yields were still maintained for the BEPT protocol (3jd’: 90% vs 90% in ABz protocol; 3id’: 85% vs 85% in ABz protocol and 60% in EGC protocol with the super-armed donor). The catalyst loading could be further reduced to 0.02 equiv without affecting the superior efficiency of the BEPT protocol over the ABz and EGC protocols (3fd’: 92% vs 80% in ABz protocol with peracetylated ABz donor and 72% in EGC protocol). Benzoyl-protected adenine 2g was not a suitable acceptor in the ABz protocol, and only 28% of 3ig was isolated; however, 60% yield of 3ig was obtained in the BEPT protocol.

Fig. 7: Examination of the practicality of the BEPT protocol.
figure 7

A Comparison of the BEPT protocol with ABz and EGC protocols (ABz o-alkynylbenzoyl, EGC ethynylcyclohexyl glycosyl carbonate). B Glycosylation of other N-nucleophiles (the chemical structures were unequivocally determined by 1D and 2D NMR). C Scalable synthesis of nucleoside drugs (the N-glycosylation steps were conducted on a gram-scale).

To further expand the acceptor substrate scope, other N-nucleophiles, including carbazoles, pyrazoles, indazoles, 1,3,4-triazoles, and asparagine derivatives, were briefly explored with the BEPT protocol (Fig. 7B)41. Under the catalysis of 0.1 equivalents of Au(I) complex, direct glycosylation of the N-heteroaromatic nucleophiles proceeded fluently, delivering 10a-i in 77–96% yields, among which 10e and 10i derived from asymmetrically substituted pyrazole and 1,2,4-triazole were obtained as mixtures of regioisomers. The chemical structures of all the obtained N-glycosides were unequivocally determined by 2D NMR (see Supplementary Information). Interestingly, synthesis of indazole-type N-glycosides by means of the BEPT protocol exhibits an evident trend that furanosyl donors give 1-N-glycosides while pyranosyl donors furnish 2-N-glycosides as determined by C3 chemical shift (130 ppm for N1- while 120 ppm N2-products) (see Supplementary Information), exquisitely complementing Taylor’s method29. The obtained N-glycosides are highly biologically relevant as the glucosyl counterpart of 10e has been used as a reference in residue analysis and risk assessment of oxathiapiprolin42, 10f-h analogs have been demonstrated cytotoxic to various cancer cell lines43, while 10i was the precursor to ribavirin, the clinically used antiviral agent with broad spectrum44. Meanwhile, the functional groups embedded in N-heteroaromatic moieties, such as fluorine, bromine, as well as the ester group, can facilitate the derivatization of these N-glycosides. Furthermore, the amide of asparagine was successfully glycosylated with BEPT donor 1 h to give glycosyl amino acid 10j in 85% yield, much higher than that of the EGC protocol20, further highlighting the advantage of the BEPT protocol. Similarly, asparagine-containing dipeptide could also be glycosylated with 1c, and 88% yield of 10k was secured. Even for the condensation between disaccharide BEPT donor 1o and asparagine-embedded tripeptide, 51% yield of 10l was still obtained.

Scalable synthesis is a widely accepted benchmark to evaluate the practicality of a synthetic method; meanwhile, low catalyst loading makes sense only in the scalable synthesis context. Thus, scale-up synthesis of nucleoside drugs was finally investigated through the BEPT protocol under reduced-catalyst loading (Fig. 7C). The condensation of 1p (1.1 g) with 2b in the presence of 0.02 equiv of Au(I) complex yielded 3pb’ efficiently (0.93 g, 83%) after silica gel-mediated product unification. Deprotection under conventional conditions transformed 3pb’ to 11a (75%, 2 steps). As a promising antitumoral drug candidate, nucleoside 11a has been proven to exhibit potent antitumor activity against a range of cancer cell lines, particularly the prostate cancer cell line Du-14545. Direct glycosylation of 2c with DHIN-bearing donor 1m was then carried out under the effect of 0.02 equiv of Au(I) catalyst. Benefited from the reaction speed-enhancing effect of HFIP, the coupling reached completion in 2 h, and 80% yield of 3mc (1.2 g) was registered after product unification via re-tert-butoxycarbonylation. The di-Boc protection greatly facilitated the photocatalyzed deoxygenation step, which was followed by silyl group deprotection and Boc’s cleavage to provide cladribine 11b (60%, 3 steps), the clinically prescribed antitumoral and immunosuppress drug46.

The practicality of the BEPT protocol was further assessed by the synthesis of intricate but biologically relevant nucleosides, angustmycin A and dJ (Fig. 8). Angustmycin A, a nucleoside cytokinin isolated from Streptomyces hygroscopicus47, proved to possess significant antibacterial48 and antitumor activities49. Glycoside dJ, a unique constituent of the DNA of Kinetoplastida and Euglina, is vital in unambiguously deciphering the function of base J (β-D-glucosyloxymethyluracil) at the molecular level, which could guide the development of therapies against diseases caused by kinetoplastid flagellates50. Equally intriguing are the novel chemical structures of angustmycin A and dJ. Rather than a common sugar motif, angustmycin A contains a 5,6-dehydrated psicofuranosyl moiety, bestowing it with a labile exocyclic double bond as well as a sterically hindered ketose anomeric carbon, which challenges the existing N-glycosylation protocols. Similarly, the β-2′-deoxyribofuranosyl N-glycosidic linkage combined with the co-existing O-glycosidic linkage also poses considerable obstacles en route to dJ by chemical synthesis11.

Fig. 8: Synthesis of augustmycin A (19), dJ (24), and dJ analog (24′).
figure 8

A Convergent synthesis of angustmycin A (N-glycosidic linkage construction after exo-glycal fabrication). B Synthesis of dJ 24 (the glycosidic linkage construction sequence is O-glycosidic linkage first and then N-glycosidic linkage). C Divergent synthesis of dJ 24 and its analog 24 (the glycosidic linkage construction sequence is N-glycosidic linkage first, followed by O-glycosidic linkage).

Although chemical syntheses of angustmycin A have been reported51, current synthetic approaches suffer from evident drawbacks, such as the use of environmentally harmful reagents, limited convergency, and unsatisfactory overall efficiency. Equipped with the BEPT protocol, an attempt to establish an alternative robust route to angustmycin A was made (Fig. 8A). Easily available diactonide-protected psicose 12 was further decorated with a 2-naphthylmethyl (Nap) group, which was followed by simultaneous selective anomeric isopropylidene removal and anomeric o-bromophenylsulfenyl group incorporation to afford 13 as a mixture of α/β-isomers (70%). Switching the resting isopropylidenyl PG in 13 to benzoyl (Bz) groups with the neighboring group-participation effect was conducted under the conventional conditions to deliver 14 (40%, 2 steps). Sonogashira reaction transformed the latent donor 14 to active BEPT donor 15 (83%). The stability of the leaving group in 15 allowed for the successive cleavage of the Nap group and iodination of the resulting OH to provide iodide 16 (95%, 2 steps). The following exo-glycal fabrication was achieved by base-mediated elimination to give 5,6-dehydrated psicosyl BEPT donor 17 (99%). Despite the challenges posed by the acid-vulnerable exocyclic double bond and the sterically hindered quaternary anomeric carbon, the coupling between 17 and 2a proceeded smoothly and stereoselectively under the catalysis of Au(I) complex at 45 °C, and the desired N-glycosides 18/18/18′′ were isolated in a combined 84% yield. Global deprotection entailed Bocs removal as well as esters saponification, converting 18/18/18′′ to angustmycin A 19 (83%, 2 steps)52. Thus, based on the BEPT protocol, a robust synthetic approach to angustmycin A, featuring exo-glycal fabrication prior to N-glycosidic linkage construction, was successfully set up (convergent synthesis). Through this approach, augustmycin A 19 was secured in an 11-step longest linear sequence (LLS) with 15% overall yield from 12.

Relying on 2′-deoxyuridine, chemical synthesis of dJ has also been accomplished53,54. Nevertheless, although bypassing the construction of the challenging 2′-deoxy-N-glycosidic linkage, the established synthetic route remains lengthy due to tedious PG manipulations. Capitalizing on the BEPT protocol and photocatalyzed sugar C-O bond editing method, concise and facile synthesis of dJ from cost-efficient starting materials was accomplished (Fig. 8B). With DCM/HFIP as the reaction medium, direct glycosylation of 5-hydroxymethyluracil 20 with BEPT donor 1b occurred preferentially on the primary OH to afford 21 (50%), along with some N,O-bisglycosylated by-product. Subsequent N-glycosylation of 21 with ribosyl BEPT donor 1m in the presence of Au(I) catalyst afforded the N,O-bisglycosylated product 22 (82%), primed for photocatalyzed deoxygenation. Under the combined effect of 4CzlPN, Ad-SH, as well as LED irradiation, 2′-deoxygenation of the ribosyl residue, initiated by the 2′-O-attached DHIN group, took place fluently to complete the assembly of the protected dJ 23 (80%). Desilylation followed by Bzs saponification realized the full deprotection of 23 to provide dJ 24 (77%, 2 steps). As such, through the approach with O-glycosidic linkage construction prior to N-glycosidic linkage, dJ was secured through a 5-step LLS in 25% overall yield, whereas the existing route provided the same product in 8% overall yield via a 7-step LLS with 2′-deoxyuridine as starting material.

An alternative approach to dJ with reversed order of glycosidic linkage construction, namely, N-glycosidic linkage first, then O-glycosidic linkage, was also devised, giving rise to not only dJ but also its 3′-deoxy analog (Fig. 8C). Taking advantage of the stability of the BEPT donors, the swap of the cyclic tetraisopropyldisiloxane-1,3-diyl (TiPDS) group for two Bzs was conducted on donor 1m. Surprisingly, notwithstanding the buffered mild desilylation conditions, DHIN group migration from 2-OH to 3-OH took place, providing 1n and 1n’ in almost equal yields (50% and 49%, respectively), which could be separated chromatographically and were separately advanced to the final products. As silylated 5-hydroxymethyluracil 25 had limited solubility in DCM, the preferred medium for glycosylation reactions, its glycosylations with 1n/1n’ were performed following the one-pot silylation/glycosylation procedure to yield N-glycosides 26 and 26 almost quantitatively. Desilylation was followed by O-glycosylation with 1b to provide 27 (78%, 2 steps from 26) and 27 (75%, 2 steps from 26). Compounds 27 and 27 were then separately subjected to photocatalyzed deoxygenation, delivering fully protected dJ 28 (80%) and its analog 28 (81%). Finally, saponification of all Bzs under basic conditions converted 28/28 to dJ 24 (99%) and its 3′-deoxy analog 24 (95%), respectively, free of by-products originating from decomposition54. Thus, following this approach, both dJ 24 and its analog 24 could be obtained through 7-step LLSs in 30% and 28% overall yield, respectively. It is noteworthy that the timing of the deoxygenation step is crucial for the success of this divergent approach. If the deoxygenation step was carried out before the O-glycosylation step, partial cleavage of the N-glycosidic linkage was observed.

In summary, thanks to the favorable activating effect on the leaving group and appropriate stabilizing effect on the key catalyst intermediate, the tert-butyl substituent vitalized a type of o-alkynylphenylthioglycosides donors, the BEPT donors. The reaction mechanism of the BEPT donor with nucleobase was investigated by model/control reactions, NMR experiments, and DFT calculations, justifying the BEPTs as viable donors in nucleoside preparation. With the stable glycosyl BEPTs as donors, a protocol for nucleosides synthesis has been established, through which direct (without resorting to nucleobase activation/sylylation), swift (the reaction time could be reduced to 10 min), and highly efficient synthesis of nucleosides has been achieved via either the PG or the solvent procedures under mild conditions (0.02–0.2 equivalents of Au(I) catalyst). Taking advantage of the stereo-controlling and deoxygenating capabilities of the DHIN group, stereoselective 2-step synthesis of 2′-deoxy nucleosides was achieved through the BEPT protocol. The beneficial effect of the HFIP was disclosed both in PG and in solvent procedures, not only dramatically speeding up the reaction rate but also allowing free uracil derivatives as viable acceptors, which dramatically streamlines the chemical synthesis of nucleosides. The underlying mechanism responsible for the rapid and regioselective synthesis of uracil-type nucleosides via the solvent procedure was elucidated by DFT calculations and NMR experiments, confirming the formation of favorable H-bond systems. Moreover, direct comparison of the BEPT protocol with the other two catalytic protocols, including the ABz and EGC protocols, was made. Other N-nucleophiles such as carbazoles, pyrazoles, indazoles, 1,3,4-triazoles, asparagine, and asparagine-embed peptides were included as viable acceptors for the BEPT protocol, and scalable synthesis of nucleosides through the BEPT protocol was achieved, leading to gram-scale preparation of chloro-substituted nucleoside and cladribine, the two highly biologically relevant nucleosides, all of which underscore the advantages of the BEPT protocol over the existing protocols. To further evaluate the practicability of the BEPT protocol, syntheses of augustmycin A (11 LLS, 15%), dJ (25%, 5 LLS, and 30%, 7 LLS), and dJ 3′-deoxy analog (28%, 7 LLS), the highly bioactive but synthetically challenging nucleosides, were accomplished. As such, the establishment of the BEPT protocol dramatically facilitates the access of nucleoside compounds, thereby considerably accelerating the process of nucleoside drug development55.

Additionally, during the synthesis of dJ and its analog, the glycosyl BEPT donor also had impressive performance in O-glycosidic linkages construction. Thus, the wide application of the BEPT protocol in O-glycosides synthesis, especially the challenging β-mannosides56 as well as the α-sialosides57, could definitely be expected. The related studies are underway in our lab, and the results will be reported in the near future.

Methods

General procedure for the N-glycosylation of purines or pyrimidines with BEPT donors

General Procedure A

A suspension of donor (1.2 equiv, 30 or 60 mM), acceptor (1.0 equiv), and freshly activated 5A molecular sieves (4 g/mmol) in dry CH2Cl2 was stirred at room temperature for 15 min under N2 atmosphere, then PPh3AuNTf2 (0.02–0.2 equiv) was added at the same temperature. The resulting reaction mixture was stirred at the same temperature until TLC showed that the coupling had reached completion. Then the mixture was filtered, and the filtrate was concentrated. The resulting residue was purified by column chromatography on silica gel to provide the corresponding product.

General Procedure B

A suspension of donor (1.2 eq, 40 mM), acceptor (1.0 equiv), and freshly activated 5A MS (4 g/mmol) in a mixed solvent of dry CH2Cl2 and HFIP (v/v = 4:1) was stirred at room temperature for 15 min under N2 atmosphere, then PPh3AuNTf2 (0.02 or 0.04 equiv) was added at the same temperature. The resulting reaction mixture was stirred at the same temperature until TLC showed that the coupling had reached completion. Then the mixture was filtered, and the filtrate was concentrated. The resulting residue was purified by column chromatography on silica gel to provide the corresponding product.

General Procedure C

A suspension of donor (1.2 equiv, 30 mM), acceptor (1.0 equiv), and freshly activated 4A MS (4 g/mmol) in dry CH2Cl2 was stirred at room temperature for 15 min under N2 atmosphere, then PPh3AuNTf2 (0.2 equiv) was added. The resulting reaction mixture was stirred at the same temperature for 4 h. Then the reaction mixture was filtered, and the filtrate was concentrated to yield the crude product, which was purified by column chromatography on silica gel to provide the corresponding product.

General Procedure D

To a sealed tube, donor (1.0 equiv), acceptor (2 equiv, 50 mM), CH2Cl2, and BSTFA (4 equiv) were added successively at room temperature under N2 atmosphere. The mixture was stirred at 80 °C until it became clear (about 8 h). The reaction mixture was then gradually cooled down to room temperature, to which 5A MS (4 g/mmol) and PPh3AuNTf2 (0.02–0.2 equiv) were added successively. The resulting mixture was stirred at the same temperature until TLC showed the completion of the reaction. Filtration and concentration under reduced pressure afforded a residue which was purified by column chromatography on silica gel to provide the corresponding product.

General Procedure E

A suspension of donor (1.0 equiv, 20 or 40 mM), acceptor (2.0 equiv), and freshly activated 5A MS (4 g/mmol) in a mixed solvent of dry CH2Cl2 and HFIP (v/v = 4:1) was stirred for 15 min at room temperature under N2 atmosphere, then PPh3AuNTf2 (0.2 equiv) was added at the same temperature. The resulting reaction mixture was stirred at the same temperature until TLC showed that the coupling had reached completion. Filtration was followed by concentration under reduced pressure to provide the crude product, which was further purified by column chromatography on silica gel to provide the corresponding product.