Microwave-Assisted Thioester-Empowered Toolkit for Modular Glycopeptide Synthesis

Abstract

Glycopeptides and glycoproteins with structural precision are valuable for functional studies and applications. Conventional couplings of glycosyl amino acids are slow, wasteful, and impractical for broad use. To address the fundamental synthetic challenges, the current work presents a new paradigm for glycopeptide synthesis, featuring a key component (thioester-functionalized glycosyl amino acid) and a comprehensive reaction system (AgSbF₆ for activation, Oxyma as an additive, DIPEA as a base, under microwave irradiation). The reaction offers rapid and clean on-resin conversion (10 min), economical reagent use (1 equiv), broad effectiveness across varied glycan structures and multiple peptide coupling sites (27 examples), and readiness to automation. Moreover, it seamlessly integrates with enzymatic glycan elaboration and protein ligation strategies, furnishing glycopeptides and glycoproteins with an increased structural complexity. Taken together, this robust and versatile platform broadens access to complex glycopeptides and glycoproteins, thereby offering a powerful entry point for functional glycoscience and biomedical discovery.

Introduction

Glycoproteins are ubiquitous and structurally complex biomolecules that play critical roles in numerous physiological and pathological processes, such as ligand–receptor binding, cell–cell communication, and immune responses. − Investigation of the functional roles of glycoproteins is largely impeded due to the difficulty in accessing pure, single glycoforms. Chemical synthesis provides glycoproteins with structural precision, thus enabling researchers to address such fundamental biological questions that would otherwise be out of reach. − Moreover, efficient and modular synthesis of glycopeptides and glycoproteins is readily transformative in glycoscience: artificial glycosylation of peptide and protein therapeutics can furnish improved potency and metabolic stability, and even new generation drugs. ,

Developments in peptide and protein chemistry have enabled the synthesis of sizable native proteins with defined modifications, − which potentiates translation into biomedical and clinical implementation. − However, the synthesis of glycopeptides and glycoproteins lags behind. Apart from the synthetic challenges in glycan moieties per se, a major hurdle is to introduce the glycan units into the protein backbone efficiently, with structural integrity. Despite the encouraging advances in direct glycosylation on peptides, − so far, the most generally available approach still relies on premade glycosylated amino acids: , through stepwise solid-phase peptide synthesis (SPPS), precise control over the glycan structure and its attachment site can be realized. It is straightforward, though, challenges remain in several ways. First of all, the coupling of glycosyl amino acids does not readily match typical SPPS protocols, either Boc-chemistry (due to the acid-lability of the glycosidic linkage) or Fmoc-chemistry (base-promoted elimination of glycans tends to occur). Moreover, owing to the increased size and complexity of glycosyl amino acids, excess amounts and repeated coupling cycles are usually needed, their economical use remaining unresolved. In addition, glycosyl amino acids are intrinsically complex so that coupling conditions may vary case-by-case, thus precluding streamlined high-throughput synthesis. Hence, a generally efficient and practical coupling protocol, particularly for the introduction of glycosyl amino acids, is highly favorable yet still lacking.

Thioesters have an ancient root in life, driving a wide range of metabolic pathways, including fatty acid, polyketide, and nonribosomal peptide syntheses. Owing to their superior reactivity and selectivity, thioesters are the most popular acyl donors in protein chemical and enzymatic syntheses. ,, Basically, peptide thioesters can undergo coupling with free amino groups, , thus directing peptide bond formation for native proteins and cyclic peptides, − even glycopeptides. − However, the inherent limitations of this reaction remained unresolved over the past decades, including long reaction time (>10 h, up to 1–2 days), need of excess thioesters (up to 7 equiv), and base-promoted side reactions such as epimerization and aspartimide formation.

Machine-driven technologies are transforming contemporary chemical synthesis. In particular, microwave-assisted automated synthesizers significantly accelerate peptide assembly, delivering speed, reproducibility, and scalability. It also contributes to tackling the “difficult sequences” containing aggregation-prone regions. We reasoned that if we start with thioester-functionalized glycosyl amino acids, to be activated under microwave irradiation, enhanced coupling efficiency should be expected; on this basis, we may reduce the reagent input and shorten the reaction time, so that side reactions can be avoided. To our delight, this simple idea proved highly successful. After extensive investigation, we managed to establish a general and robust protocol for glycopeptide synthesis, termed Microwave-Assisted Thioester-Empowered glycopeptide Toolkit (MateGPT) (Figure b). By exploiting the unique reactivity of thioesters under microwave irradiation (50 °C), the MateGPT method features clean and fast reaction (quantitative conversion in 10 min), extremely low stoichiometry (1 equiv of thioester-functionalized glycosyl amino acid and each reagent), broad effectiveness (among varied glycan sizes/configurations/linkages and peptide residues even with less reactivity), and readiness to fit automated platforms. An illustrative example is the streamlined construction of a synthetic library of α-dystroglycan (α-DG) glycoforms, enabled by the MateGPT standard protocol. Furthermore, MateGPT readily integrates with state-of-the-art technologies, i.e., enzymatic glycan elaboration and native chemical ligation, to furnish glycopeptides and glycoproteins with increased structural complexity. Collectively, this method offers a modular, economical, and versatile platform, opening the door to scalable and routine access to complex glycopeptides and glycoproteins, thus, paving the way for functional studies of protein glycosylation events.

1.

Prior strategies and our synthetic design for glycopeptide synthesis. (a) Conventional solid-phase synthesis of glycopeptides. (b) Our strategy: Microwave-assisted thioester-empowered glycopeptide synthesis.

Results and Discussion

Reaction Development

Our study initiated with α-dystroglycan (α-DG), a transmembrane cell-adhesion receptor that bridges the extracellular matrix and the intracellular actin cytoskeleton. , O-Mannosylation, consisting of diverse core (M1, M2, and M3) and their extending glycan structures, is found to be crucial to the function of α-DG, whereas abnormal O-mannosylation accounts for a group of clinical disorders termed dystroglycanopathy. To look into the intricate structural differences and their functional implications, our primary target was to obtain a panel of homogeneous glycosylated variants of α-DG (317–334), bearing O-mannosylation at Thr317. Although Pro318 at the coupling site may pose additional challenges due to its steric and electronic constraints, we took it as a stringent model for testing our method.

To economize on the use of costly glycosyl amino acids, we tested the coupling with merely Fmoc-L-Thr­(α-D-Man­(Ac₄))–OH (1A) (Figure ). However, the conventional SPPS coupling conditions fell short due to either poor coupling efficiency or base-promoted side reactions, including β-elimination product 5 and intramolecular acyl transfer product 6 (entries 1–8; selected reaction traces shown). While the use of certain additives can enhance the coupling yield and alleviate side reactions (entry 5 vs entry 6), depletion of base significantly diminished the reaction yield (entry 7). This dilemma prompted us to develop alternative strategies.

2.

Reaction setup: On-resin amidation between the glycosyl amino acid (1A) or the corresponding thioester (2A) and the peptidyl resin 3a. ^a 1A (1 equiv) or 2A (1 equiv), 0.04 M, anhydrous DMF. The apparent yield was obtained based on the UV absorption of peptide peaks in HPLC traces of the crude reaction at 220 nm. ^b Reaction was performed at microwave 70 °C. Selected HPLC traces of the reaction mixture are shown.

We next explored a thioester-based approach (entries 9–15), since it is shelf-stable per se (see Supporting Information for detailed stability investigation), electrically neutral, and of moderate polarity that eases large-scale preparation and purification. We reasoned that its soft activation via a thiophilic reagent such as Ag­(I) ion may circumvent the pitfalls of base-promoted activation. To test this hypothesis, the corresponding thioester Fmoc-L-Thr­(α-D-Man­(Ac₄))-STol (2A, 1 equiv) was subjected to coupling with peptide resin 3 using AgNO₃ for activation and 1-oxo-2-hydroxydihydrobenzotriazine (HOOBt) as an additive. The reaction outcome was significantly improved, producing the desired glycopeptide 4Aa in 79% yield within 3 h (entry 9). Notably, shortening the reaction time to 1.5 h still delivered 75% conversion with even a lower level of side products (entry 10), reinforcing that the kinetics of Ag­(I)-promoted thioester activation highly favored productive coupling. Considering the limited solubility and high UV absorption of HOOBt, which complicated HPLC analysis and purification, we examined other additives as well. While HOBt and Oxyma were more soluble, neither of them afforded better efficiency (entries 11 and 12).

To optimize the activation process, various Ag­(I) salts were also investigated (Table S2). Notably, AgSbF₆ offered immediate solubilization in DMF and excellent transformation: glycopeptide 4Aa was obtained in 90% yield within 1.5 h (entry 13). Extension of the reaction time up to 6 h did not deteriorate the yield or selectivity (entries 14–15), highlighting the robustness and mildness of the reaction system.

To further improve the reaction profile and streamline the synthesis, our goal was to accelerate the reaction rates, minimize reagent use, and adapt the method to automated platforms. Hence, we transferred the reaction to a microwave-assisted peptide synthesizer, as microwave irradiation is essentially helpful and outperforms conventional methods at room temperature or via conductive heating. To our delight, this approach is highly effective (entries 16–20). Under microwave irradiation (50 °C, 10 min), the reaction employing 2A, AgSbF₆, Oxyma, and DIPEAeach at 1.0 equivdelivered the desired product 4Aa in nearly quantitative yield (entry 16). Raising the temperature to 70 °C offered no additional benefits or deleterious effects (entry 17), and lowering AgSbF₆ to catalytic amounts severely diminished the conversion (entry 18). Consistent with previous findings, excess DIPEA increased β-elimination (entry 19), while its absence hindered reaction completion (entry 20). As a control study, microwave-assisted coupling using HATU and DIPEA (entry 21) led to moderate reaction outcome, thus highlighting the uniqueness of the MateGPT protocol as a superior surrogate.

To gain insight into the reaction course, a series of mechanistic experiments were carried out. First, glycosyl amino acid thioester 2A was examined under MateGPT conditions in the absence of the peptidyl resin. Quantitative formation of an Oxyma-derived active ester intermediate Int-1 was observed (Scheme a and Figure S89 in the Supporting Information). Subsequently, treatment of this intermediate Int-1 with the peptidyl resin 3b under microwave irradiation afforded the glycopeptide product 4Ab in excellent yield (Scheme b; see also Figure S90 in the Supporting Information). Together, these results support a two-step reaction process (Scheme c). The glycosyl amino acid thioester I is activated by silver hexafluoroantimony to generate Oxyma ester III, which then undergoes rapid coupling with peptide IV through amide bond formation to furnish the desired glycosyl peptide V. Thus, it suggests a dual role of thioester, not only as a protected form of glycosyl amino acid but also as a reactive precursor in the presence of Ag­(I). The reaction is highly accelerated under microwave conditions, enabling quantitative transformation within a few minutes.

1. Mechanistic Studies and Proposed Mechanism of MateGPT (a–c).

Epimerization is a longstanding concern in glycopeptide synthesis, particularly for glycosylated serine derivatives. To rigorously evaluate the stereochemical integrity of the MateGPT protocol, we selected glycosyl serine-derived thioesters as representative examples. Briefly, O-mannosylated Fmoc-L-Ser­(α-D-Man­(Ac₄))-STol (2B) and its D-isomer (2C), as well as D- and L-Ser thioesters bearing O-GlcNAcylation (2P and 2Q), were synthesized and subjected to standard MateGPT conditions with peptidyl resin 3b (Figure ). To our delight, in all of these cases, the corresponding glycopeptides (4Bb, 4Cb, 4Pb, and 4Qb) were obtained as the exclusive products. HPLC analysis of each crude reaction mixture revealed no detectable epimerization (<1%), thus confirming that the MateGPT protocol maintains stereochemical fidelity during the peptide coupling process (see also Figures S92–S93 in the Supporting Information for more details).

3.

(a, b) Epimerization study under standard MateGPT condition. HPLC traces of the crude reactions at 220 nm are presented.

Reaction Scope

Unlike N-linked glycosylation, which typically involves a conserved pentasaccharide core attached to Asn within the consensus Asn-Xaa-Ser/Thr motif, O-glycosylation exhibits high structural diversity and lacks a universal sequon. This heterogeneity presents a major synthetic challenge and, at the same time, serves as a rigorous system for evaluating the scope and robustness of glycopeptide synthesis methods.

To assess the generality of the MateGPT strategy, we explored its performance across a broad range of peptide sequences (3a–3m) and thioester-functionalized glycosyl amino acids bearing structurally distinct glycans (2A–2M) (Table ). Unlike conventional SPPS coupling conditions that require at least 2 equiv of glycosyl amino acids and repeated coupling cycles, MateGPT achieves quantitative conversion with 1 equiv of each reagent in 10 min, without detectable side products. Notably, it was highly effective for peptide substrates with sterically hindered β-branched residues (Val and Ile) and less nucleophilic Pro at the coupling site, affording glycopeptides 4Ac, 4Ad, and 4Ae in excellent yields. Amino acids susceptible to side reactions, including Tyr, His, and Trp, remained chemically intact throughout the reaction, leading to the clean formation of glycopeptides 4Af–4Ah. Furthermore, Cys residues with various protections such as Trt, Acm, or StBu were well tolerated, furnishing products 4Ab, 4Ai, and 4Aj without cleavage or rearrangement. For oxidation-prone methionine-containing peptides, the reaction was conducted under an inert atmosphere to successfully yield glycopeptide 4Ak in 90% yield. Peptide fragments from natural glycoproteins such as α-DG (322–334) and Notch-1 receptor (466–475) reacted smoothly to give the desired O-mannosyl glycopeptides 4Al and 4Am in high yield, further manifesting the utility of the current method.

1. General Scope Across Peptide Sequences and Diverse Glycan Moieties .

^aStandard condition of MateGPT: peptidyl resin (3a–3m, 1 equiv), thioester-functionalized glycosyl amino acids (2A–2M, 1 equiv), AgSbF₆ (1 equiv), Oxyma (1 equiv), DIPEA (1 equiv) in anhydrous DMF (0.04 M), microwave 50 °C, 25 W, 10 min. All reactions were conducted under the standard condition unless otherwise specified. The crude purity was estimated based on the UV absorption of peptide peaks in HPLC traces of the crude reactions at 220 nm. The cumulative isolation yield was calculated based on resin loading and shown in parentheses.

The method also demonstrated high versatility with respect to the glycan structure. Disaccharide-furnished thioesters 2D (β-D-lactoside) and 2E (β-D-primeveroside) were efficiently coupled, yielding glycopeptides 4Db and 4Eb, respectively, as exclusive products. Rare glycosidic linkages, including O-linked phenolic glycosides (Tyr, 4Fb) and S-linked thioglycosides (Cys, 4Gb), were well accommodated as well. More impressively, the success also extended to N-linked glycopeptide synthesis: thioester 2H, bearing a β-GlcNAc at Asn, delivered 4Hd at a conserved N-glycosylation site. It is particularly encouraging that no aspartimide side product was apparently formed under the MateGPT condition, whereas additional precautional measures are usually required for N-linked glycopeptide synthesis such as incorporation of a pseudoproline dipeptide moiety in the peptide backbone. , Additionally, glycopeptide 4Hd stands as a valuable precursor to endoglycosidase-catalyzed transglycosylation to give various glycoforms.

To serve our primary goal, a panel of α-DG (317–334) glycoforms were prepared under the MateGPT condition, bearing α-D-mannopyranoside (4Aa), β-D-glucopyranoside (4Ia), β-D-galactopyranoside (4Ja), β-D-xylopyranoside (4Ka), 2-acetamido-2-deoxy-β-D-glucopyranoside (4La), and 2-acetamido-2-deoxy-β-D-galactopyranoside (4Ma), respectively. It is worth mentioning that even in the absence of further optimization, each reaction took place at the Pro site in quantitative conversion, highlighting the potential of MateGPT as a general and high-throughput platform for glycopeptide and glycoprotein synthesis.

Streamlined Synthesis of α-DG Glycopeptides with Multiple O-Mannosylation

With the notion that varied glycosylation patternsaltering glycosylation sites and/or glycan structurescan significantly affect protein and peptide properties, we pursued to verify the capacity of MateGPT through the streamlined synthesis of glycopeptides with multiple glycosylation sites bearing individual glycans (Figure a). First of all, Fmoc-L-Thr­(α-D-Man­(Ac₄))-STol 2A was selected for a brief demonstration. To our expectation, the iterative coupling of 2A under a standard MateGPT protocol (1 equiv, 50 °C microwave irradiation) proved highly effective. In this way, the consecutive introduction of O-mannosylation at Thr322 and Thr317 was smoothly achieved (noted as Reactions i and ii, HPLC traces and ESI-MS characterization in Figure b), with a slight level of Ac loss detected cumulatively. Upon one-pot global removal of Ac and Fmoc groups, the final product 8 was obtained in 19% overall yield (4.1 mg isolated by HPLC, on a 0.01 mmol scale, based on the initial loading of the resin).

4.

Streamlined synthesis of α-DG glycopeptide (317, 334) with multiple glycosylations. (a) Synthetic scheme of O-mannosyl glycopeptides 8 and 13, respectively. MateGPT indicates the standard reaction condition under microwave 50 °C for 10 min. (b) HPLC traces (220 nm) and ESI-MS characterization of major reactions (i-iv) and purified products 8 and 13. Peak * denotes loss of one Ac group during the microwave irradiation. Peak # denotes loss of one Man residue in the mass spectrometer. Peak ∧ denotes loss of one GlcNAc residue and Peak ∧∧ denotes loss of two GlcNAc residues due to fragmentation in the mass spectrometer.

More encouragingly, glycopeptide 13, bearing both Core M1 and Core M2 O-mannosyl glycans, was also successfully prepared through the MateGPT protocol. With reaction (iii) between peptide fragment 3l and Core M1 disaccharyl thioester 9, glycopeptide 10 was obtained as the exclusive product; after SPPS extension, the next round of coupling (reaction (iv)) employing Core M2 trisaccharyl thioester 11 brought the formation of product 12 (Figure ). Subsequent acyl removal and Fmoc deprotection led to the final product 13 in a 20% overall yield (isolated by HPLC). Remarkably, a scaled-up synthesis afforded comparable efficiency (30.6 mg, 23% overall yield), highlighting the scalability of the approach. It was noted that on such a preparative scale, extended TFA treatment should be avoided to minimize the loss of glycan units. Considering the structural complexity of the target product, the robustness and effectiveness of the current approach is manifested.

Enzymatic Modification of α-DG Glycopeptides Furnishing Complex Glycan Structure

To further expand the structural complexity, glycopeptide 13, bearing native Core M1 and Core M2 O-mannosyl glycans, was an excellent substrate for enzymatic glycan elaboration (Figure ). Following our established protocol, , stepwise enzymatic diversification was performed. To our delight, clean conversion and a decent isolation yield were obtained for each step. First of all, three galactose residues (Gal) were introduced concomitantly in the presence of UDP-Gal via a β(1→4)-linkage, affording glycopeptide 14 carrying LacNAc-terminated Core M1 trisaccharide and Core M2 pentasaccharide, respectively, in quantitative conversion and 71% yield (isolated by HPLC). Starting there, the purified glycopeptide 14 served as an acceptor for enzymatic α(2→6)-sialylation, affording fully sialylated glycopeptide 15 exclusively, in 65% yield (isolated by HPLC).

5.

Enzymatic glycan modification leading to α-DG glycopeptides (317 and 334) with complicated glycan structure. (a) Synthetic scheme. (b) HPLC traces (220 nm) of crude reaction mixture with the formation of desired products; ESI-MS characterization of key intermediates and products. Peak # denotes loss of one Gal residue and Peak ∧ denotes loss of one Neu5Ac residue due to fragmentation in the mass spectrometer.

These encouraging results motivated us to challenge a cascade reaction in a one-pot multienzyme manner. Notably, well-furnished glycopeptide 16 bearing the native Neu5Ac-α(2→3)-Gal-β(1→4) motif at each glycan chain was obtained in a 52% overall isolation yield. Considering the steric hindrance posed by both linear Core M1 and branched Core M2 glycans and the close proximity of the glycosylation sites, such an achievement is quite remarkable. Together, these results underscore the utility of our strategy in enabling the modular assembly of glycopeptides with multiple O-glycosylation sites and elaborate glycan structures, through a combination of chemical and enzymatic transformations.

MateGPT and One-Pot NCL Led to O-Fucosylated EGF Domain of Notch-1 Receptor

To further demonstrate the versatility of the MateGPT protocol, we targeted the synthesis of Notch-1 receptor, a key component of the evolutionarily conserved Notch signaling pathway. , Human Notch-1 receptor contains tandem epidermal growth factor (EGF)-like repeats on its extracellular domain, many of which are glycosylated with unusual forms of O-glycans, such as O-fucose (Fuc) glycans. Glycosylation is found to be essential for the regulation of Notch signaling in a vast variety of developmental processes. , To enable functional investigations, a site-defined O-fucosylation was required. While our previous work on stereoselective glycosylation provided a facile preparation of O-fucosylated amino acid building blocks, , how to convert these glycosyl moieties into glycopeptides precisely and efficiently remains to be explored. To this end, the established AgSbF₆-promoted MateGPT protocol readily offers a convenient way.

To optimize the overall synthetic efficiency, retrosynthetically, the EGF12 domain was disconnected into a glycopeptide hydrazide , 19 (452–475) and a peptide fragment 20 (476–488). Following the general MateGPT procedure employing Fmoc-L-Thr­(α-L-Fuc­(2-Ac-3,4-Bz))-STol 17, quantitative conversion of on-resin coupling led to peptidyl glycopeptide 18 (Figure , Reaction (v)); subsequent SPPS and global deprotection yielded glycopeptide hydrazide 19 (43% overall yield after HPLC purification). It was then activated in the presence of acetylacetone (acac) and 4-mercaptophenylacetic acid (MPAA), followed by one-pot native chemical ligation with peptide 20, thus affording O-fucosylated EGF12 domain 21 in 56% overall yield. This result again demonstrates the power and practicality of the MateGPT protocol for glycopeptide and glycoprotein synthesis.

6.

Synthesis of the O-fucosylated EGF12 domain via MateGPT and one-pot NCL. (a) Synthetic scheme. (b) HPLC traces (220 nm) and ESI-MS characterization of crude reaction mixture with the formation of desired products and characterization of purified products. Peak # denotes loss of one Fuc residue due to fragmentation in the mass spectrometer.

Conclusion

In summary, we successfully established a practical protocol for glycopeptide synthesis, employing the glycosyl amino acid in its thioester form. Upon activation in the presence of AgSbF₆ as a promoter, Oxyma as an additive, and DIPEA as a base, an efficient on-resin coupling reaction can be effected during solid-phase peptide synthesis. The current protocol features microwave-aided rapid reaction (ca. 10 min), extremely economical use of building blocks (as low as 1 equiv of each reagent), high efficiency (single-shot, peak-to-peak conversion), and general applicability (to various peptide sequences and glycan moieties). As it is highly adaptable to the automated microwave-based peptide synthesizer, the simplified workflow facilitates routine access to structurally diverse and complex glycopeptides. An illustrative example here is the streamlined synthesis of α-DG glycopeptides with multiple O-mannosylation, namely Core M1 and Core M2 glycans, respectively; increased complexity was achieved via subsequent enzymatic glycan elaboration. Moreover, it leads to the smooth synthesis of EGF12 domain of Notch-1 receptor with O-fucosylation, a crucial modification to modulate the Notch signaling pathway. Collectively, this work expands the synthetic repertoire for glycopeptides and provides a scalable platform for generating structurally defined glycoforms. In this way, MateGPT is expected to accelerate the development of glycopeptide libraries for functional glycomics, biomedical translation, and therapeutic discovery.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.5c01264.

• Detailed experimental procedures, characterization data, copies of NMR spectra for all new compounds (PDF)

#.

Y.Y., M.Z. and M.-H.X. contributed equally to this work. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.