Introduction

Carbohydrates play indispensable roles as natural building blocks and in therapeutics, diagnostics, and functional foods1,2,3. Glycomimetics, including C-, N-, and S-glycosides, often retain the biological functions of their natural O-glycoside counterparts while exhibiting distinct chemical properties, making them valuable tools for probing glycan function and facilitating drug discovery4. Among these, C-glycosides have attracted sustained interest across chemical, biological, and medical disciplines owing to their characteristic C–C glycosidic bonds, which impart hydrolytic and metabolic stability compared to traditional C–O linkages5,6,7. Within this subclass, aryl C-glycosides are distinguished by the direct attachment of aryl groups at the anomeric C-1 position of sugar rings. This structural motif has attracted significant attention due to its unique chemical properties, druggability, and diverse biological activities, including applications in cancer agents8, antibiotics9, and antidiabetic drugs10. Of particular note are sterically hindered aryl C-glycosides 1, which feature bulky aryl aglycones and are prevalent in numerous bioactive natural products and clinically relevant glycomimetics (Fig. 1a)5. Representative examples include Saptomycin B 2, isolated from Streptomyces sp. HP530, which exhibits potent in vitro and in vivo antitumor activities11. Another notable case is the α-C-mannosylation of tryptophan moiety 3 in glycoproteins, a rare and sterically demanding C-glycosylation occurring as a post-translational modification in TSR family proteins12. Additionally, spiro C-arylated glycopyranosides such as Papulacandin D 4 demonstrate antibiotic and antifungal properties13. Fused cyclic aryl C-glycosides, including Bergenin 5, widely distributed in medicinal plants, possess broad biological profiles encompassing antitumor, antiviral, and antifungal properties14. The glycosylated angucycline Gilvocarcin M 6 displays activity against Gram-positive bacteria15. More recently, Enavogliflozin 7, a sterically hindered aryl C- glycoside, was identified as an important sodium-glucose cotransporter 2 (SGLT-2) inhibitor for treating type 2 diabetes patients with impaired renal function16. Sterically hindered C-glycosidic linkages are also a characteristic structural feature within the angucycline family. For example, Marmycin A 8 exhibits potent anticancer activity, attributed in part to its sterically congested C-glycosidic linkages17.

Fig. 1: The background and development of sterically hindered aryl C-glycosides.
Fig. 1: The background and development of sterically hindered aryl C-glycosides.
Full size image

a Representative examples of sterically hindered aryl C-glycoside in natural products and pharmaceuticals. b Background and general strategies for the synthesis of aryl C-glycosides. c Work design: Ligand-enabled next-generation stereospecific glycosyl cross-couplings to access hindered aryl C-glycosides. P protected groups, R substituent groups, Lg leaving groups, L ligands, X halogen, Ar aryl group, Ar’ hindered aryl group, GCC glycosyl cross-couplings.

Given their broad significance as key glycomimetics, aryl C-glycosides have attracted sustained interest as synthetic targets for stereoselective anomeric arylation6,7,18,19,20,21,22,23. Established chemical methods for C-glycosylation can be broadly classified into three principal strategies based on the nature of the reactive glycosyl intermediates involved (Fig. 1b): (i) Friedel−Crafts-type reactions of electron-rich arenes with glycosyl electrophiles, or direct phenol O-glycosylation followed by stereoselective O → C rearrangement, both proceeding via glycosyl oxocarbenium ions24,25,26,27 (pathway I); (ii) radical glycosyl cross-coupling reactions, with or without transition metal catalysis, involving glycosyl radicals (pathway II)28,29,30,31,32,33,34; and (iii) transition-metal-catalyzed glycosyl cross-coupling (GCC) via stereospecific transmetalation of glycosyl anions (pathway III)35,36. Despite significant advancements, achieving highly controllable, chemoselective, and stereoselective aryl C-glycosylation—particularly for sterically hindered aryl C-glycosides—remains a considerable and ongoing challenge. Classical substitution methods involving glycosyl oxocarbenium ions encounter notable obstacles, including reliance on stoichiometric Lewis acid promoters, preference for electron-rich arenes, and difficulties in managing chemoselectivity between C- and O-glycosylation pathways. Controlling anomeric stereochemistry for sterically hindered aryl C-glycosides is especially demanding due to the combined effects of the saccharide framework, C2 substituents on sugar rings, and the electronic and steric profiles of bulky aryl groups—all of which exacerbate stereoselectivity issues at the highly congested anomeric center (Fig. 1b, pathway I)6,24,25. Recently, radical C-glycosylation has gained prominence as a powerful strategy for the synthesis of aryl C-glycosides30,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58, offering broad functional group tolerance, good anomeric selectivity, and compatibility with unprotected sugar substrates. In radical-mediated C-glycosylation, the saccharide backbone strongly governs anomeric control, with substrate-dictated stereoselectivity often observed, typically favoring α-stereoselectivity. However, the stereodivergent synthesis of both optically pure α- and β-glycosides remains a formidable challenge in carbohydrate chemistry, yet is essential for the systematic investigation of glycan structure–function relationships38,52,59,60,61. Moreover, the electronic and steric properties of the aryl coupling partners can markedly impact the α/β selectivity62,63, rendering the stereoselective construction of sterically hindered aryl C-glycosides particularly challenging (Fig. 1b, pathway II). Transition metal-catalyzed glycosyl cross-coupling (GCC) reactions, which unite glycosyl nucleophiles with aryl electrophiles, have emerged as a powerful strategy for constructing aryl C-glycosides (Fig. 1b, pathway III)35. In this context, we have successfully developed highly stereospecific Pd-catalyzed cross-coupling reactions between both α- and β-anomers of glycosyl stannanes and aryl halides, enabling the modular and predictable synthesis of stereochemically pure α- and β-aryl C-glycosides59,60,64. This strategy showcases broad substrate scope, excellent functional group tolerance—including compatibility with fully unprotected sugars—and consistently high stereospecificity. A key to this success lies in the stereoretentive transmetalation step that preserves the anomeric configuration regardless of the steric or electronic properties of both the glycosyl donors and aryl electrophiles. This principle has been elegantly demonstrated in palladium-catalyzed stereoretentive intramolecular glycosylations, enabling efficient access to both cis- and trans-fused cyclic C-glycosides without erosion of optical purity61. However, while the stereoretentive nature of these transformations ensures precise anomeric control, the application of sterically hindered aryl electrophiles in intermolecular glycosyl cross-coupling remains highly challenging, with such substrates frequently exhibiting severely diminished reactivity or complete inertness under established conditions (Fig. 1b, pathway III)60,61.

Motivated by unresolved synthetic challenges and the notable bioactivity of sterically hindered aryl C-glycosides, and guided by the high selectivity and reliability achieved in our preliminary work59,60, we aimed to develop a next-generation stereospecific glycosyl cross-coupling platform using glycosyl stannanes to access these sterically demanding structures. The rapid evolution of transition metal-catalyzed cross-coupling reactions has been intimately linked to a deeper understanding of ligand steric and electronic effects and their influence on metal reactivity65,66,67,68,69. In particular, the successful of Csp2-Csp3 Stille glycosylation reaction has relied on electron-deficient phosphine ligands, which enhance reactivity by generating more electrophilic palladium(II) species, compensating for the low nucleophilicity of glycosyl stannanes, and facilitating efficient transmetalation59,60,70. The key to achieving efficient synthesis of sterically hindered aryl C-glycosides lies in the rational design of ligands tailored for these challenging aryl electrophiles. Such an effective ligand should possess: (i) electron-withdrawing properties to create more electrophilic Pd(II) species, thereby facilitating transmetalation in the presence of bulky aryl electrophiles; and (ii) suitable steric effects, where bulkiness aids reductive elimination but excessive steric hindrance may hinder oxidative addition with bulky electrophiles and the subsequent transmetalation of Pd(II) species bearing hindered aryl groups.

Here, we report a next-generation, ligand-enabled, stereospecific glycosyl cross-coupling for glycosyl stannanes with sterically bulky aryl electrophiles 14, providing efficient and robust access to structurally demanding aryl C-glycosides (Fig. 1c). This efficient approach allows for precise stereocontrol over the anomeric configuration of both α- and β-anomers derived from various saccharides. Through systematic modification of the first-generation ligand JackiePhos L1, we identified two uniquely effective variations of Jackiephos biaryl phosphine ligands, L10 and L11, bearing methoxy and isopropoxy substituents at the 2’-and 6’-positions of the “lower” aryl ring, respectively, specifically designed to facilitate C–C reductive elimination. These ligands outperform the original JackiePhos system, particularly in promoting the coupling of ortho-monosubstituted and ortho-disubstituted sterically hindered aryl halides. The versatility and robustness of this refined glycosyl Stille cross-coupling protocol were demonstrated across more than 65 examples, featuring both protected and unprotected mono- and oligosaccharides, and a diverse range of bulky aryl electrophiles, including bioactive natural products and pharmaceutical agents. Computational analysis reveals that the underexplored ligand L11 favors reductive elimination to forge sterically hindered C–C bonds over β-methoxy elimination, opposite to the first-generation ligand L1, consistent with experimental observations. Natural bond orbital (NBO) analysis further revealed that Pd in Int3 (L11) carries a slightly higher NBO charge (+0.18) than in Int3′ (L1, +0.16). The more electron-rich Pd center in Int3′ better stabilizes the β-methoxy elimination transition state, favoring this pathway, whereas the electron-deficient Pd in Int3 facilitates C–C bond formation, promoting reductive elimination.

Results and discussion

Progress in transition metal-catalyzed cross-couplings has been largely propelled by the development of increasingly effective ligands. To identify a suitable ligand for glycosyl C(sp3)−C(sp2) cross-coupling involving sterically hindered aryl electrophiles, we integrated insights from established ligand effects and mechanistic studies of Stille-type couplings71,72. As a model, we tested the reaction of (2,3,4,6-tetra-O-benzyl-β-D-glucopyranosyl)tributylstannane 16 and the sterically hindered methyl 2-bromobenzoate 14a under previously reported Stille glycosyl cross-coupling reactions (Fig. 2)59,60,61. JackiePhos L1, a biarylmonophosphine ligand featuring strongly electron-withdrawing P-bound 3,5-bis(trifluoromethyl)phenyl groups, has been widely used as an optimal ligand for stereospecific Stille glycosyl C(sp3)−C(sp2) couplings of aryl electrophiles59,60,61,64,70,73. However, it proved less effective for the title reaction, affording only 15% of coupling product 17. As anticipated, substituting the P-bound electron-withdrawing (3,5-bis(trifluoromethyl)phenyl) groups in L1 with electron-donating cyclohexyl substituents, as in BrettPhos L2, completely abolished reactivity. This result is consistent with prior findings that electron-deficient phosphine ligands are crucial for enabling Stille C-C couplings59,60,70, thereby ruling out the use of electron-rich phosphines for promoting the oxidative addition of sterically hindered electrophiles. Tris(3,5-bis(trifluoromethyl)phenyl)phosphine L3 delivered product 17 in 21% yield but with extensive formation of by-products via β-O elimination (17-1, 79%) and β-H elimination (17-2, 49%). Recognizing that transmetalation is the rate-limiting step in Stille glycosyl C(sp³)−C(sp²) cross-coupling reactions59, particularly for bulky aryl electrophiles, we hypothesized that more electrophilic and less sterically demanding Pd(II) intermediates would facilitate critica transmetalation step. Maintaining electron-withdrawing substituents on the phosphorus atom while removing the two methoxy groups on the “upper” aryl ring A of the biaryl backbone, we systematically modified the “lower” aryl ring B by varying substituents at the 2′, 4′, and 6′ positions (see Fig. 2 for numbering). Ligands L4-L8 showed progressively improved yields (24% to 70%) alongside suppression of by-product 17-1 formation. This finding indicates that, contrary to prior reports59,74, the methoxy groups on the “upper” aryl ring, particularly at the C3 position, are not indispensable for sterically hindered cross-coupling reactions. Earlier studies suggested that the methoxy group located in the ortho position to phosphorus is critical for maintaining catalyst stability through coordination with the O-bound Pd (II) center, which is challenging to form due to the increasing steric interference from the bulky aryl ring. Compared to JackiePhos L1, its analogue L4 (lacking the two methoxy groups on the “upper” aryl ring A) provided only a modest yield improvement (15% vs. 24%). To advance this objective, we designed a ligand, L5, bearing hydrogen atoms at the 2’, 4’, and 6’ positions of the “lower” aryl ring B to reduce the steric hindrance. This modification led to a notable increase in the yield of 17 to 62%, compare to 24% obtained with L4. When 2’ and 6’ substituent groups were replaced with alkoxy groups, such as methoxy L6, cyclohexyloxy L7, and isopropoxy L8, the yields improved to 70%, 39%, and 46%, respectively, compared to L4. These results suggest that an electron-rich “lower” aryl ring might benefits the reaction by strengthening coordination to the Pd(II) center via the C-bound Pd (II)75. However, excessive steric demand of electron-donating groups like cyclohexyloxy (L7) and isopropoxy (L8) may partially offset these benefits. This conclusion also extends to ligands L9-L12, which feature an “upper” aryl ring A and ring C identical to JackiePhos L1. These ligands showed improved yields ranging from 66% to 82%, along with a significant reduction in β-H elimination. Comparing ligand pairs L5 vs. L9, L6 vs. L10, and L7 vs. L11 further established that while methoxy groups on ring A are not strictly necessary, their presence consistently enhances catalytic performance. This effect is particularly pronounced when R1 is i-Pr, with the yield increasing significantly from 46% to 73%, likely due to the C3 methoxy substituent acting as a flexible coordination site that stabilizes the electrophilic and the sterically hindered aryl-Pd (II) center B, even when O-bound Pd(II) intermediates are disfavored sterically. Interestingly, relocating the methoxy group from the C6 to the C5 position on the “upper” aryl ring enhanced the performance of L13 over L1, though this positional adjustment slightly decreased the efficiency of L14 relative to L10. Further modification of the “lower” aryl ring yielded ligands L15 (fused cyclic alkoxy groups) and L17 (carbazolyl group). Despite their differing steric demands, both ligands furnished comparable yields (64% and 71%, respectively), indicating no further advantage beyond the trends already observed. Finally, we investigated variations in the “upper” ring of the biaryl moiety. Four ligands (L16, L18, L19, and L20), each bearing the optimized “lower” aryl ring, were synthesized, incorporating different groups in the upper ring: indenyl L16, indole L18, imidazole L19, and pyrazole L20. These variations afforded the desired product 17 in yields of 39%, 48%, 0%, and 52%, respectively. These results underscore the critical role of the upper aryl ring in modulating catalytic performance. Notably, all reactions across this ligand set consistently furnished the β-anomer exclusively, even with sterically hindered aryl electrophiles, highlighting the high stereocontrol, robustness, and reliability of the glycosyl cross-coupling protocol. Based on these evaluations, ligands L10 and L11 were identified as optimal for further applications and mechanistic investigations.

Fig. 2: Optimizing ligands for next-generation stereospecific glycosyl cross-coupling.
Fig. 2: Optimizing ligands for next-generation stereospecific glycosyl cross-coupling.
Full size image

Standard reaction conditions: 14a (0.05 mmol), 16 (0.075 mmol), Pd2(dba)3 (2.5 mol%), ligand (10.0 mol%), CuCl (300 mol%), KF (200 mol%), 1,4-dioxane (1.50 mL), 110 °C, 48 h, N2, The yields and stereochemical outcome were determined by 1H NMR analysis of crude reaction mixtures. a isolated yield.

A detailed understanding of oxidative addition palladium (II) complexes is central to elucidating how ligands influence metal coordination and catalytic reactivity in cross-coupling reactions76,77. Biaryl monophosphine ligands, in particular, have proven indispensable in enabling C–C bond formation65. Accordingly, the synthesis and structural characterization of these Pd(II) intermediates remain key to advancing ligand design and mechanistic insight. For dialkylbiaryl monophosphine ligands, extensive studies have established both C-bound OA1 and O-bound OA2 isomers, supported by crystallographic data75,78,79,80,81,82,83. In contrast, although C-bound OA3 and O-bound OA4 species derived from diarylbiaryl monophosphine ligands have been frequently proposed74, their structures remain poorly defined due to the lack of crystallographic evidence. This structural ambiguity may have contributed to the relatively limited development and application of diarylbiaryl ligands in catalysis. Motivated by this, we investigated the structures of diarylbiaryl monophosphine ligand-based Pd(II) complexes, aiming to explore whether unconventional binding modes might be accessible beyond the established paradigms (Fig. 3a).

Fig. 3: Synthesis and characterization of diarylbiaryl monophosphine-based oxidative addition complexes (OACs).
Fig. 3: Synthesis and characterization of diarylbiaryl monophosphine-based oxidative addition complexes (OACs).
Full size image

a The background and development of oxidative addition Pd complexes. b Synthesis and characterization of OA5. c Synthesis and characterization of OA6. d Synthesis and characterization of OA7. e Synthesis and characterization of OA8. R substituent groups, X halogen, Ar aryl group.

To investigate the structural behavior of diarylbiaryl monophosphine ligands, we synthesized oxidative addition Pd(II) complexes using hindered ortho-methylbromobenzene 14b as the electrophile and analogues of JackiePhos L11 and Jackiephos L1 as ligands. Crystallization from the mixture of pentane and DCM afforded the desired oxidative addition complexes (OAC) OA5 and OA6 in 78% and 82% yields, respectively (Fig. 3b, c). Unexpectedly, X-ray analysis revealed dimeric Pd(II) species bridged by two bromide atoms, rather than the anticipated monomeric C-bound OA3 or O-bound OA4 structures. We initially hypothesized that the unusual dimeric geometry observed might arise from the use of a sterically hindered electrophile. However, performing the reaction with a non-hindered substrate, ethyl 4-bromobenzoate 14c, yielded the same dimeric Pd(II) species OA6, indicating that electrophile sterics alone do not dictate dimer formation (Fig. 3d). During the course of this work, the Buchwald group reported the synthesis and NMR characterization of a Pd(II) complex bearing a 2-naphthyl group and supported by a JackiePhos/CPhos hybrid ligand81. Although no crystal structure was provided, the authors proposed a C-bound OA3-type monomer, inconsistent with the bromide-bridged dimeric species we consistently observed. This was particularly intriguing, as the hybrid ligand shares the same A and C aryl rings as JackiePhos L1 and L11, suggesting that L1-and L11-based Pd(II) complexes might also form well-defined species amenable to crystallization. To examine this, we repeated the synthesis using the hybrid ligand. Remarkably, single crystals suitable for X-ray diffraction were obtained, which revealed the same dimeric Pd(II) species bridged by two bromide rather than the proposed OA3 monomer (Fig. 3e). Notably, dissolution of OA5 in CD₂Cl₂ gave a mixture of bromide-bridged dimeric Pd(II)–OA complexes and O-bound OA2 isomers, as revealed by NMR, indicating that the bromide-bridged dimers are the more stable species, coexisting in dynamic equilibrium with the less stable O-bound OA2 isomers in solution. In contrast, dissolution of OA6, OA7, and OA8 in solvent exclusively yielded the bromide-bridged dimeric Pd(II)–OA complexes (see Supplementary Note 3.1 for experimental details). With well-defined oxidative addition palladium(II) complexes OA5 and OA6 in hand, we evaluated their catalytic performance in the reaction between glucosyl stannane 16 and methyl 2-bromobenzene 14b under standard conditions (see Supplementary Note 3.2 for experimental details), affording the desired product in 99%, and 32% yield, respectively. Together, these findings indicate that the formation of dimeric Pd(II) species is an inherent structural feature of diarylbiaryl monophosphine ligands, largely independent of the steric or electronic nature of the aryl electrophile or the P-bound substituents. More importantly, this challenges a prevailing assumption and clarifies a key structural ambiguity, informing future ligand design and expanding the scope of biarylmonophosphine ligands in catalysis.

To gain deeper insight into and accurately predict the impact of different ligands on reaction yield and chemoselectivity (β-H and β-O elimination), we developed multivariate linear regression models that correlate ligand properties with the yields of various products and byproducts (Fig. 4). Initially, experimental data were collected for 20 different ligands. Ligand characteristics were then quantified using a (PdL) reaction model, based on the above crystal structures obtained from our own experiments (Fig. 3). These descriptors included electronic parameters such as NBO parameters (natural charges on Pd, P, C1, C2, and C3), the Wiberg bond index (WBI_Pd_P), and frontier orbital energies (HOMO, LUMO). Steric parameters included percent buried volume (%Vbur) for Pd and P, sterimol parameters (bmin, bmax, L along Pd to P), as well as bond lengths (Pd–P, P–C) and bond angles (Pd–P–C). To construct and validate the models, the dataset was randomly divided into training and test sets at an 80:20 ratio. All three models showed strong linear correlations, with R² values of 0.85, 0.89, and 0.90, respectively. Although based on a modest dataset of 20 ligands, the MLR model shows strong correlations (R² = 0.85–0.90) and good agreement with test set predictions. Cross-validation (80:20 split) confirms its robustness, and undoubtedly, a larger dataset could further improve generality. The predicted yields for the four ligands in the test set were in close agreement with the experimental results, with the exception of the 17-2 product from ligand L4, which exhibited a noticeable deviation. The resulting models revealed distinct structure–property relationships for each product. For product 17, the yield was positively correlated with higher HOMO energy and C3_NBO charge, and negatively correlated with higher C1_NBO, bmin_Pd_P, and bmax_Pd_P values. For product 17-1, increased C1_NBO and bmin_Pd_P values were associated with higher yields, whereas increased C3_NBO and longer P–C3 bond length had a negative effect. In the case of product 17-2, higher values of HOMO energy, C1_NBO and C2_NBO, and longer P–C1 bond length were associated with lower yield. These findings demonstrate that by fine-tuning ligand electronic and steric properties—such as electron density distribution, frontier orbital energies, and Pd–P coordination characteristics—it is possible to simultaneously optimize reaction efficiency and selectivity, thereby enabling the targeted enhancement of desired product yields.

Fig. 4: Multivariate models linking ligand structure to product distribution in catalytic reactions.
Fig. 4: Multivariate models linking ligand structure to product distribution in catalytic reactions.
Full size image

1. Data collection. 2. Property description. 3. Model building. NBO, natural bond orbital.

Kinetic studies revealed a first-order dependence on 16, the palladium catalyst, and KF, indicating that all three components participate directly in, or precede, the rate-determining step (Fig. 5a). In contrast, a zero-order dependence was observed for the aryl iodide, suggesting that the oxidative addition is rapid and not rate-limiting. To gain further mechanistic insight, detailed DFT calculations were performed (Fig. 5b). To optimize computational efficiency, the OBn group was replaced with OMe, and Bu₃Sn was replaced with Me₃Sn. The computed Gibbs free energy profile for the reaction catalyzed by JackiePhos-derived biaryl phosphine ligand L11 is presented. The iodine-bridged dimeric Pd(II) species int1, formed upon oxidative addition, was identified as the resting state of the catalytic cycle and taken as the reference point (0.0 kcal/mol). The catalytic cycle proceeds through transmetalation and reductive elimination steps to deliver the major product. Oxidative addition of the aryl iodide regenerates the dimeric Pd(II) complex, thereby completing the cycle. Dissociation of the dimeric Pd(II) species requires an energy input of 10.2 kcal/mol. In the presence of KF, int2 undergoes stereoretentive transmetalation with stannane 16 via a six-membered cyclic transition state ts-164, forming intermediate int3 and significantly lowering the activation barrier to 26.7 kcal/mol. This transmetalation step is identified as turnover-limiting, consistent with the kinetic data. An alternative stereoretentive transmetalation pathway proceeding through a four-membered cyclic transition state (ts-1′) was calculated to require a significantly higher barrier of 37.8 kcal/mol, rendering it unfavorable. From int3, the reaction can proceed via C-C reductive elimination to afford the major product. Alternatively, in the presence of CuCl and F⁻, an E2-type β-methoxy elimination pathway can generate a side product59. Reductive elimination via ts-2 from int3 occurs with a modest activation barrier of 7.8 kcal/mol, yielding the desired aryl C-product and regenerating the active Pd(0) catalyst. Subsequently, oxidative addition of the aryl iodide onto the Pd(0) species proceeds via transition state ts-3, with a low energetic span of 2.2 kcal/mol, forming an intermediate analogous to int2. Notably, the β-methoxy elimination pathway via ts-2a (in the presence of CuClF⁻) has an overall energy barrier of 10.0 kcal/mol, which is 2.2 kcal/mol higher than that for C–C bond-forming reductive elimination from Int3. These computational results suggest that the indentifed phosphine ligand L11 promotes the C–C reductive elimination pathway, thereby favoring formation of the desired aryl C-product over side-product formation.

Fig. 5: Reaction energy profile of the Pd-catalyzed Stille coupling of hinder 2-bromotoluene using L1 and L11 ligands.
Fig. 5: Reaction energy profile of the Pd-catalyzed Stille coupling of hinder 2-bromotoluene using L1 and L11 ligands.
Full size image

a Kinetics studies. b Density functional theory (DFT) studies.

In contrast, when the ligand L11 is replaced with the first-generation JackiePhos L1, the computational results show the opposite trend. From intermediate int3′, the C–C bond-forming reductive elimination proceeds via transition state ts-2’ with an activation barrier of 9.8 kcal/mol, whereas the competing β-methoxy elimination pathway via ts-2a′ requires only 7.8 kcal/mol, rendering it more favorable. This trend aligns well with experimental observations, where β-methoxy elimination is the dominant pathway with ligand L1. Structurally, no significant geometric differences were observed between intermediates int3 and int3′ that could clearly account for this reversal in selectivity. To further probe the electronic factors underlying this divergence, we compared the natural bond orbital (NBO) charges on the palladium center in both intermediates. The calculated NBO charge on Pd in int3 is +0.18, while it is slightly lower at +0.16 in int3′. This subtle difference in electron density may influence the relative activation barriers for the two competing pathways. The slightly more electron-rich Pd center in int3′ (L1) may better stabilize the developing negative charge in the β-elimination transition state, thus lowering its energy relative to C–C reductive elimination. Conversely, the more electron-deficient Pd in int3 (L11) may favor the reductive elimination step by facilitating bond formation with the aryl group. While these electronic effects are modest, they are consistent with the observed reactivity trends and may contribute to the distinct selectivity observed with the two ligands.

Having identified optimal ligand structures, we next evaluated their performance in the cross-coupling of sterically hindered aryl electrophiles. Each bulky (hetero)aryl halide was tested with the first-generation ligand JackiePhos L1 and the underexplored second-generation ligands L10 and L11 to benchmark their efficacy in aryl C-glycosylation. As shown in Fig. 6a broad array of sterically hindered aryl bromides and iodides, including ortho-substituted arenes and heterocycles, underwent smooth coupling with L10 or L11, delivering the desired C-glycosides in good to excellent yields with exclusive anomeric control. In stark contrast, L1 afforded low yields or failed to promote product formation altogether. We systematically probed the impact of steric and electronic substituents using ortho-substituted aryl halides as a representative platform. ortho-Substituents such as ester 18, methyl 21-27, 2-ethoxy-2-oxoethyl 28, hydroxymethyl 29, isopropyl 30, benzo 31, phenyl 32, phenoxy 33, benzyloxy 34, acetoxy 35, hydroxy 36, acetylamino 37, benzoyl 38, formyl 39, cyano 40, and nitro groups 41 were all well tolerated, affording the corresponding C-glycosides in moderate to excellent yields. Notably, several substrates that were unreactive or less reactive using L1 as ligand, including those bearing hydroxymethyl, isopropyl, phenyl, benzyloxy, acetylamino, benzoyl, cyano, and nitro groups, exhibiting markedly enhanced reactivity with L10 and L11. The transformation demonstrates broad functional group compatibility, including tolerance toward acidic phenolic hydroxyl, hydroxymethyl, acetamido, and formyl groups. Even in the context of more sterically congested 1,5-disubstituted aryl halides, fully unprotected glycosyl stannanes could be efficiently coupled to deliver products 42–44 in 60–79% yield using L10 or L11, while L1 gave only 12–21% yield under otherwise identical conditions.

Fig. 6: Scope of bulky electrophilic coupling partners.
Fig. 6: Scope of bulky electrophilic coupling partners.
Full size image

Standard reaction conditions: 14 (0.05 mmol), 16 or 17 (0.075 mmol), Pd2(dba)3 (2.5 mol%), ligand (10.0 mol%), CuCl (300 mol%), KF (200 mol%), 1,4-dioxane (1.50 mL), 110 °C, 48 h, N2, isolated yields for L10, 1H NMR yields for L11 and Jackiephos L1. P protected groups, R substituent groups, X halogen, Ar’ hindered aryl group.

Encouraged by these promising results, we next explored the substrate scope of glycosyl stannanes using sterically demanding electrophiles such as ethyl 2-bromobenzoate 14a and ethyl 4-bromo-3-methylbenzoate 14e as representative models. As shown in Fig. 7, a broad range of (un)protected glycosyl stannanes efficiently participated in the cross-coupling, delivering the corresponding sterically hindered aryl C-glycosides in moderate to excellent yields while consistently maintaining exclusive anomeric selectivity for either the β- or α-isomer. Glucosyl stannanes bearing common groups (OBn, OPiv, OLev, OBz, and OTIPS) exhibited excellent compatibility under the standard conditions with ligand L10, affording products 20 and 47-48 in isolated yields of 63%-93%. Fully disarmed peracetylated and perbenzoylated 2-deoxy-D-glucosyl stannanes smoothly reacted to afford the corresponding products 49 and 50 in 79% and 72% yield, respectively. Moreover, structurally diverse glycosyl stannanes derived from a series of monosaccharides, including D-galactose 51, 2-deoxy-D-glucose (52, 53), D-mannose 54, reversed D-galactose 5564, reversed D-mannose 56, L-rhamnose 57, underwent efficient cross-coupling to furnish the desired hindered aryl C-glycoside. Notably, achieving both optically pure α- and β-isomers under a single set of conditions remains rare in C-glycosylation38,59,60,64. The second-generation glycosyl cross-coupling platform reliably yielded both α- and β-isomers (for example, 52 and 53) with complete stereocontrol even in the presence of sterically hindered aryl electrophiles. The ability to access both optically pure α- and β−2-deoxy aryl C-glycosides (52 and 53) highlights the advantage of this method in overcoming the longstanding reliance on C2-neighbouring group participation for stereocontrol in C-glycosylation reactions. Furthermore, unprotected or partially unprotected aryl C-glycosylation has historically posed a formidable challenge due to difficulties in achieving both high chemo- and stereoselectivity42,53,59,60. Encouragingly, the indentified ligands, in combination with hindered aryl electrophiles, enabled efficient coupling of unprotected glycosyl stannanes to deliver partially and fully unprotected aryl C-glycosides (59–62) in good yields and with high stereoselectivity. The successful synthesis of fully unprotected, sterically hindered aryl C-glycosides 61 and 62 provides compelling evidence that transition metal-catalyzed glycosyl cross-coupling via glycosyl nucleophiles remains the most robust and arguably the only general strategy for accessing both optically pure anomers, irrespective of the sugar structure or the steric and electronic properties of the aryl coupling partner. Significantly, oligosaccharide stannanes 63 and 64 were also viable coupling partners without modification of the standard conditions, delivering aryl C-disaccharides in 67% and 60% yields, respectively. Anomeric configurations were assigned by analysis of 1H NMR spectra of the crude reaction mixtures, based on the characteristic 3J(HH) coupling constants of the H1 proton. These signals typically appeared at 3.81 − 5.25 ppm range (CDCl3), with J values of 8.80−9.40 Hz for 1,2-trans isomers and 4.20−6.20 Hz for 1,2-cis isomers.

Fig. 7: Scope of anomeric stannanes.
Fig. 7: Scope of anomeric stannanes.
Full size image

Standard reaction conditions: 14 (0.05 mmol), 45 (0.075 mmol), Pd2(dba)3 (2.5 mol%), L10 (10.0 mol%), CuCl (300 mol%), KF (200 mol%), 1,4-dioxane (1.50 mL), 110 °C, 48 h, N2, isolated yields. P protected groups, R substituent groups, X halogen, Ar’ hindered aryl group.

Late-stage glycodiversification of drug molecules offers a direct and promising strategy for modulating pharmaceutical properties, including solubility, metabolic stability, and target specificity4. However, this process poses significant challenges, as it requires a delicate balance between chemoselectivity and stereoselectivity, all while accounting for the electronic properties and steric hindrance of the drug molecules84. Motivated by the high chemo- and stereoselectivity of our developed glycosyl cross-coupling platform and its compatibility with sterically hindered substrates, we sought to apply this strategy to the late-stage glycodiversification of bioactive molecules and approved drugs using fully unprotected D-glucosyl stannanes (Fig. 8). Using this method, a series of aryl C-glycosylated pharmaceutical derivatives, including Gemfibrozil 67, Clofibrate 68, Fenofibrate 69, Isoxepac 70, Flurbiprofen (71 and 72), Naproxen 73, and Xanthotoxin 74, were synthesized via a concise and highly selective sequence. Key to this process was the regioselective C–H halogenation (bromination or iodination) of aryl scaffolds85,86,87, followed by palladium-catalyzed glycosyl cross-coupling with full unproteced glycosyl stannane 18. The reaction displayed remarkable chemoselectivity in halogenated substrates (e.g., 68, 69, 71, 72), with no erosion of stereochemical integrity even in structurally complex molecules such as Naproxen 73. Notably, electron-rich and sterically hindered substrates (e.g., 68–70, 74) typically challenging for cross-coupling reactions underwent efficient glycosylation under standard conditions. As anticipated, by tuning the site of halogenation, glycosylation patterns could be precisely modulated, enabling programmable, site-selective late-stage glycodiversification. This precise late-stage C-glycosylation protocol provides a reliable, general, and robust strategy for directly introducing glycan motifs into complex drug-like molecules and natural products, expanding access to various glycosylated analogues for biological evaluation and pharmaceutical optimization.

Fig. 8: Late-stage sterically demanding glycodiversification of bioactive molecules and drug scaffolds.
Fig. 8: Late-stage sterically demanding glycodiversification of bioactive molecules and drug scaffolds.
Full size image

Standard reaction conditions: 65 (0.05 mmol), 18 (0.10 mmol), Pd2(dba)3 (2.5 mol%), L10 (10.0 mol%), CuCl (300 mol%), KF (200 mol%), 1,4-dioxane (1.50 mL), 110 °C, 48 h, N2, isolated yields. bSaponification reaction is carried out after the reaction is completed. R substituent groups, X halogen.

1,2-Annulated sugars have attracted considerable attention owing to their diverse structures and potent biological activities88. Among them, fused cyclic aryl C-glycosides represent a valuable yet underexplored subclass, largely due to the pronounced steric congestion surrounding the anomeric center. While established cross-coupling strategies are well-suited for acyclic aryl C-glycosides6, their direct application to the synthesis of fused cyclic aryl C-glycosides has been severely hampered by steric constraints, often compromising stereoselectivity89. Previously, we reported the stereospecific intramolecular glycosyl cross-coupling strategy for the construction of both cis- and trans-fused cyclic C-glycosides61. However, this approach required additional manipulations, including the pre-installation of a 2-iodobenzyl group at the C2-OH position of the glycosyl donor, thereby limiting its overall synthetic efficiency. To facilitate the sterically demanding coupling, the method relied on an intramolecular design, high palladium catalyst loading (10 mol%) and prolonged reaction times (72 h). To address these longstanding challenges and demonstrate the practical utility of our sterically hindered glycosyl cross-coupling platform, we devised a direct and highly stereospecific strategy for the one-step synthesis of fused cyclic aryl C-glycosides. This approach capitalizes on the efficient Stille glycosylation of glycosyl stannanes 77 bearing a free C2-OH group and bulky aryl halides 14-1 containing an ortho-ester substituent, enabling simultaneous formation of C–C glycosidic and C–O bonds in a single operation. As shown in Fig. 9a, a variety of common glycosyl stannanes coupled smoothly with methyl 2-bromobenzoate 14a, affording the desired fused cyclic aryl C-glycosides 79–83 in isolated yields of 29–57%. Moreover, other sterically hindered aryl halides 14–1, featuring substituents such as chloro 84, aldehyde 85, naphthyl 86, ester 87, and methyl 88, were also well tolerated under the optimized conditions, delivering the corresponding products in moderate yields. Notably, the anomeric selectivity of all annulated C-glycosides is governed by stereoretentive transmetalation, with α-anomeric stannanes affording exclusively the α-product and β-anomeric stannanes yielding the β-product.

Fig. 9: Synthesis of fused cyclic aryl C-glycosides, drugs and natural products.
Fig. 9: Synthesis of fused cyclic aryl C-glycosides, drugs and natural products.
Full size image

a One-step synthesis of fused cyclic aryl C-glycosides. b Total synthesis of both anomers of Enavogliflozin. c Total synthesis of Chrysin 8-C-β-D-glycosides. P protected groups, Ar’ hindered aryl group, DMF N,N-dimethylformamide; DCM dichloromethane, green balls represent reaction steps.

Navogliflozin, a next-generation SGLT2 inhibitor for type 2 diabetes, features a highly sterically hindered aryl C-glycosidic bond, a structural element critical to its pharmacological efficacy and safety90. A previously reported synthesis of compound 7 began with the reaction between an aryl lithium reagent and lactone 76. Following this key C–C bond-forming step, four additional transformations were required to reconstruct the anomeric configuration and adjust protecting groups88. Notably, the method relied heavily on reactive and flammable reagents, posing operational challenges and yielding exclusively the β-anomer. In contrast, our approach enables the direct coupling of sterically hindered aryl bromides with fully unprotected glycosyl stannane 18 and its α-anomer 18-α in a single step, delivering the corresponding β- and α-anomers in 96% and 68% yield, respectively, with excellent stereoselectivity (Fig. 9b). This streamlined route not only simplifies the synthetic sequence but, crucially, provides access to both anomers, facilitating more comprehensive structure activity relationship studies.

Chrysin 8-C-β-D-glucosides are natural flavone aryl C-glycosides isolated from the herbs of Psidium guajava, featuring a sterically congested β-C-glycosidic linkage and exhibiting broad biological activities, including anti-diabetic, antibacterial, antiviral, antioxidant, anti-inflammatory, and anti-tumor effects89. Given its pronounced steric hindrance and electron-rich aromatic core, chrysin 8-C-β-D-glucoside was chosen as a model to further showcase the robustness of our next generation of glycosyl Stille cross-coupling conditions (Fig. 9c). The synthesis began with methylation of hydroxyl groups of Apigenin 89 to afford 90 in 86% yield, followed by regioselective bromination at C-8 to give 91 in 82% yield. Subsequent intermolecular Stille coupling of 91 with C2-OH unprotected β-D-glucosyl stannane 45n proceeded under standard conditions, furnishing 92 in 13% yield but with exclusive anomeric control. The low yield observed for the cross-coupling of substrate 91 arises from its combined steric hindrance and electron-rich character, which hinders both oxidative addition and transmetalation. Increasing the palladium loading does not improve the yield and instead leads to significant β-OBn elimination from glycosyl copper intermediates, indicating that these intermediates cannot efficiently undergo transmetalation and that steric hindrance in transmetalation likely represents the rate-limiting factor. Final global debenzylation with BCl₃ provided chrysin 8-C-β-D-glucoside 93 in 81% yield.

To illustrate the advantages of our second-generation glycosyl Stille cross-coupling reaction in constructing sterically hindered aryl C-glycosides, we chose electron-rich, sterically demanding 2-bromo-4-methoxyphenol 94, which bears multiple potential reactive sites, along with 4-methoxyphenol 98, as model substrates to systematically investigate three different approaches involving glycosyl anions, glycosyl cations, and glycoside radicals (Fig. 10). As expected, both optically pure α- and β-isomers of glucosyl stannanes coupled smoothly with 2-bromo-4-methoxyphenol 94, affording the corresponding β-aryl C-glycosides 95 and α-aryl C-glycosides 96 in yields of 58% and 15%, respectively, with exclusive anomeric control and no detectable O-glycoside 99 formation (Fig. 10a). This outcome highlights the excellent chemo- and stereoselectivity of our method. In contrast, classical Friedel–Crafts-type glycosylations of electron-rich arenes with glycosyl electrophiles via glycosyl cations, typically Lewis acid-catalyzed, often lack chem- and stereoselectivity24,25. For example, anomerically pure α- or β-O-glucosyl trichloroacetimidate 97 reacted with 4-methoxyphenol 98 under catalytic TMSOTf conditions furnished only O-glycosides 99 with similar α/β ratios (1:1.26 and 1:1.30), rather than the desired α-hydroxyaryl C-glycosides 95 or 96 (Fig. 10b). This result reflects the oxocarbenium-ion-mediated mechanism of Schmidt glycosylation, in which the initial anomeric configuration is largely erased (Fig. 10b).[15a] More recently, nickel-catalysed reductive coupling of glycosyl halides with aryl halides via glycoside radicals have emerged as an alternative route to aryl C-glycosides63. To assess this approach for sterically hindered C-glycosides, the Ni/tBu-Terpy/Zn system was tested with glucosyl bromide 100 and 2-bromo-4-methoxyphenol 94, but no desired products 95 or 96 were obtained (Fig. 10c). These comparisons further emphasize the robustness, broad compatibility, and high selectivity of our Pd-catalyzed strategy. Furthermore, although the distinct advantages of second-generation ligands (L10 and L11) over first-generation ligands L1 with sterically hindered electrophiles were firmly established in the substrate scope, it was necessary to confirm their performance with non-hindered substrates. As shown in Fig. 10d, coupling of model substrate 16 with bromobenzene under identical conditions afforded the product in 80%, 76%, and 80% yields with L1, L10, and L11, respectively. These results demonstrate that the second-generation ligands are broadly effective for both sterically hindered and unhindered systems, thereby justifying their replacement of the first-generation Stille glycosyl cross-coupling reactions35,36.

Fig. 10: Comparison of our method with established approaches.
Fig. 10: Comparison of our method with established approaches.
Full size image

a Our method. b glycosyl cations method, c glycoside radicals method. d Investigation of the second-generation ligands in unhindered systems.

In summary, we developed a highly efficient and stereospecific Pd-catalyzed glycosyl cross-coupling protocol for the synthesis of sterically hindered aryl C-glycosides, enabled by two underexplored biaryl monophosphine ligands (L10 and L11). Featuring electron-withdrawing P-bound 3,5-bis(trifluoromethyl)phenyl groups and 2′,6′-methoxy (L10) or isopropoxy (L11) substituents, these ligands outperform first-generation JackiePhos (L1), showing enhanced reactivity with bulky aryl electrophiles and overcoming long-standing limitations in hindered aryl C-glycosylation. Computational analysis shows that the underexplored ligand L11 favors reductive elimination to form sterically hindered C–C bonds over β-methoxy elimination, opposite to the first-generation ligand L1, consistent with experimental results. NBO analysis further reveals that Pd in Int3 (L11) carries a slightly higher charge than in Int3′ (L1); the more electron-rich Pd in Int3′ stabilizes the β-methoxy elimination transition state, whereas the electron-deficient Pd in Int3 promotes C–C bond formation via reductive elimination. Crystal structure analysis revealed that bromide-bridged dimeric Pd(II) species are intrinsic to diarylbiaryl monophosphine ligands—challenging a longstanding assumption and resolving a structural ambiguity relevant to ligand design. The developed method shows broad applicability, accommodating diverse (un)protected sugars and sterically demanding aryl partners with consistently high chemo- and stereoselectivity. Overall, this work provides a timely solution to a long-standing challenge in the stereoselective synthesis of sterically hindered aryl C-glycosides, advances glycomimetic drug discovery, and offers deeper insights into biaryl monophosphine ligand design, thereby expanding the scope of Pd-catalyzed cross-couplings.

Methods

General procedure for cross-Coupling of glycosyl stannanes with hindered aryl halides. Inside an argon-filled glovebox, an oven-dried 3 mL vial was charged with anomeric stannane (1.5–2.0 equiv), aryl halide (1.0 equiv), Pd2(dba)3 (2.5–5 mol%), L10/L11/L1 (10–20 mol%), CuCl (300 mol%), KF (200%) and anhydrous 1,4-dioxane (1.5 mL). The vial was capped and taken outside of the glovebox. The resulting mixture was placed into a pre-heated (110 °C or 140 °C) aluminium block and stirred for the indicated period of time, cooled to rt, filtered through a pad of Celite, and concentrated. 1H NMR spectra was recorded using this mixture to evaluate diastereoselectivity. The crude material was purified by column chromatography on SiO2.