A facile chemoenzymatic synthesis of SARS-CoV-2 glycopeptides for probing glycosylation functions

Abstract

Glycosylation plays important roles in SARS-CoV-2 infection. We describe here a facile chemoenzymatic synthesis of core-fucosylated N-glycopeptides derived from the SARS-CoV-2 Spike protein and their binding with glycan-dependent neutralizing antibody S309 and human lectin CLEC4G. The synthetic glycopeptides provide tools for further functional characterization of viral glycosylation.

Graphical Abstract

Structurally well-defined synthetic SARS-CoV-2 glycopeptides provide useful probes for characterizing the glycan binding specificity of lectin and neutralizing antibody

The outbreak of COVID-19, a severe acute respiratory disease caused by SARS-CoV-2, has become a global health crisis.¹ According to the World Health Organization (WHO), it caused more than 124 million confirmed cases and over 2.7 million deaths by the end of March 2021. SARS-CoV-2 invades host cells through the spike protein-mediated binding, via its receptor-binding domain (RBD), to human angiotensin-converting enzyme 2 (hACE2) as the host receptor.² The S-protein and particularly the RBD are the primary target of current vaccine strategies aiming to elicit neutralizing antibodies to block the binding of spike protein to ACE2 receptor.³ The SARS-CoV-2 spike protein is heavily glycosylated with 22 conserved N-glycosylation sites and several potential O-glycosylation sites for each S-protein protomer (Fig. 1A).^4–8 Similar to other viral proteins including HIV-1, SARS-CoV and MERS, the heavy glycosylation usually provides shielding of immunogenic epitopes to evade host immune response.⁹ Recent studies have indicated that SARS-CoV-2 glycosylation also plays other active roles beyond the immune evasion. For example, the two conserved N-glycans located at N331 and N343 sites of the RBD have been shown to be essential for viral infectivity, deletion of which drastically reduced the infectivity;¹⁰ and the N-glycans at the N165 and N234 sites are critical to modulate and stabilize the RBD’s active conformations required for its efficient binding to host receptor hACE2;¹¹ In addition, It has been implicated that glycan-mediated interactions may play a role in tissue differentiation through binding with glycan specific human lectins located in different parts of human body.¹² For example, human lectin CLEC4G (hCLEC4G) has been found to interact with SARS-CoV and Ebola virus through the viral surface glycoproteins.¹³ Recently, It has been reported that hCLEC4G significantly reduces SARS-CoV-2 infections, presumably by binding to the N343 N-glycan in the RBD, which is mostly core-fucosylated and terminated with GlcNAc moiety, to block hACE2.¹⁴ On the other hand, the N-glycans are also an essential part of the epitopes of several neutralizing antibodies. For example, S309, a monoclonal antibody that was isolated from a SARS-CoV patient and could neutralize both authentic SARS-CoV and SARS-CoV-2 viruses, recognizes a conserved core-fucosylated N-glycan at N343 site as an essential component of the neutralizing epitope.¹⁵ Despite the diverse functions, however, a detailed understanding of the structure-function relationship of SARS-CoV-2 glycosylation is hampered by the remarkable structural heterogeneity and complexity of S-protein glycosylation. Thus, there is a high demand of structurally well-defined, homogeneous glycopeptides/glycoproteins for mapping the neutralizing epitopes, the lectin specificity, and the functional roles of respective N-glycans on the spike protein.

Figure 1.

N-Glycosylation of SARS-CoV-2 Spike Protein. (A) 22 N-glycosylation sites on spike protein. (B) Glycoform abundance observed across the receptor-binding domain (RBD). (C) the interaction between RBD and receptor ACE2; the sequence and structure of the RBD cyclic glycopeptide with the two conserved N-glycosylation sites. (D) Synthetic N-glycopeptides carrying structurally well-defined N-glycans.

The receptor binding domain (RBD) (aa319-537) carries two conserved N-glycans at the N331 and N343 glycosylation sites (Fig. 1A), which are complex-type N-glycans with up to 98% core-fucosylation and with very low level of sialylation (Fig. 1B).⁴ Recently, Wang and co-workers have reported the first semi-synthesis of several homogeneous glycoforms corresponding to the RBD using a combination of chemical synthesis and expressed protein ligation.¹⁶ The glycosylated RBD has been used for binding analysis with S309. However, the work has not provided the naturally dominant core fucosylated glycoform found at the N343 glycosylation site, which appears to be essential for S309 recognition.¹⁵ We report in this paper a highly convergent chemoenzymatic synthesis of a series of homogeneous N-glycopeptides corresponding to the sequence (aa325-366) of the glycosylated loop of the SARS-CoV-2 RBD (Fig. 1C, I) and a preliminary study on their use for characterizing the glycan-binding specificity of antibody S309 and human lectin CLEC4G. Glycopeptides carrying the naturally dominant core fucosylated complex type N-Glycan (G0F) on N343 (Fig. 1D, II) or N331 (Fig. 1D, III) or both sites (Fig. 1D, IV) were designed. A library of glycopeptides carrying different N-glycans at N343 site, including G2F (Fig. 1D, V), G0 (Fig. 1D, VI) and Man5 (Fig. 1D, VII) were also designed for structure-activity relationship studies.

Synthesis of complex glycopeptides carrying large glycans remains a challenging task due to the complexity in multistep chemical synthesis. We sought to apply a chemoenzymatic method to construct the designed glycopeptides, using endoglycosidase mutants to transfer N-glycans to an Asn-linked GlcNAc or fucosylated GlcNAc moiety at a predetermined glycosylation site, which would be more flexible and efficient than pure chemical synthesis for constructing diverse N-glycopeptides because of its high convergence and efficiency.^{17, 18} The synthesis started with the preparation of peptide carrying GlcNAc at the N343 glycosylation site through automated solid-phase peptide synthesis (SPPS) based on Fmoc chemistry on a Rink Amide AM resin (Scheme 1). Briefly, GlcNAc-Asn building block was introduced into the N343 site during SPPS, and an alkyne tag was placed at the N-terminus for future conjugation purpose. The peptide was then cleaved from the resin by treatment with Cocktail R (TFA/thioanisole/EDT/anisole, 90/5/3/2, v/v) with simultaneous global deprotection of the acid-labile protecting groups. Removal of the remaining O-acetyl groups from the protected GlcNAc moiety at N343 with 2.5% aqueous hydrazine also led to simultaneous cyclization to form a disulfide bond between C336 and C361, yielding GlcNAc peptide precursor 1. The GlcNAc peptide precursors carrying a GlcNAc at N331 (2) and at both N331 and N343 (3) were obtained following the same strategy. The peptide with no GlcNAc (4) was synthesized by SPPS as a control.

Scheme 1.

SPPS of SARS-CoV-2 RBD S325-S366 (glyco)peptides carrying GlcNAc at predetermined N-glycosylation sites and efficient core-fucosylation catalyzed by AlfC E274A.

With the GlcNAc peptide precursors in hand, we next focused on construction of the glycopeptides by installing the core fucose followed by enzymatic sugar chain elongation. We have recently described an α-fucosidase mutant (AlfC E274A) that could transfer a fucose moiety from simple α-fucosyl fluoride to GlcNAc-peptides to form the corresponding core fucosylated GlcNAc-peptides.¹⁹ Adopting this method, we synthesized the Fucα−1,6-GlcNAc-peptides (5, 6 and 7) by reaction of the GlcNAc-peptides with α-fucosyl fluoride under the catalysis of AlfC E274A (Scheme 1). For sugar chain elongation, we used an EndoF3 glycosynthase mutant, N165A, which has been shown to be efficient to transfer an N-glycan from its activated glycan oxazoline precursor to a core-fucosylated GlcNAc moiety to form complex type glycopeptides and glycoproteins.²⁰ We found in this study that EndoF3-N165A could also transfer a degalactosylated complex type N-glycan oxazoline (8) to the fucosylated GlcNAc-peptides (5, 6 and 7), to provide glycopeptides 11, 15 and 16, respectively (Scheme 2). Thus, the chemoenzymatic synthesis of the core fucosylated glycopeptides was successfully achieved through SPPS followed by two efficient enzymatic glycosylations. The G2F glycopeptide (12) was obtained using a galactosylated glycan oxazoline (9) as the donor substrate in a similar manner. The synthesis of glycopeptide 13 carrying a non-fucosylated G0 N-glycan at N343 was accomplished by an enzymatic reaction between GlcNAc-peptide (1) and glycan oxazoline 8 under the catalysis of EndoM N175Q²¹ Similarly, the glycopeptide (14) carrying a high mannose N-glycan was synthesized by the EndoM N175Q catalyzed reaction of GlcNAc-peptide 1 and the Man₅GlcNAc-oxazoline (10) (Scheme 2). The purity and identity of the synthetic glycopeptides (4 and 11-16) were confirmed by HPLC and ESI-MS analysis (Supplementary Fig. S4 and S8–13). In addition, we also synthesized the corresponding biotinylated N-glycopeptides (18-24) by copper (I)-catalyzed alkyne-azide cycloaddition reaction between the alkyne-glycopeptides (11-16) and an azide-functionalized biotin (Scheme 2). These biotin-tagged glycopeptides will be useful for ELISAs and for cell sorting analysis. It should be pointed out that the facile chemoenzymatic synthesis described here is highly convergent and flexible for making a range of structurally well-defined SARS-CoV-2 glycopeptides which would be otherwise difficult to obtain by pure chemical synthesis.^{16, 22}

Scheme 2.

Chemoenzymatic synthesis of glycosylated SARS-CoV-2 RBD 42-mer peptides

The access to these homogeneous glycopeptides makes it possible to probe the glycan-binding specificity of glycan-dependent neutralizing antibodies and human lectins involved in SARS-CoV-2 spike protein recognition. As a preliminary study, we first tested the binding of the synthetic glycopeptides with SARS-CoV-2 neutralizing antibody S309. The cryo-EM analysis of S309 in complex with SARS-CoV-2 spike protein has shown that S309 recognizes specifically a core-fucosylated N-glycan at N343 together with a segment of peptide.¹⁵ We performed ELISA analysis of the binding between S309 and the glycopeptides (11-16) or SARS-CoV-2 spike protein trimer (His Tag, Acrobiosystems Inc.), which were coated on the plates. As expected, the S-protein trimer was shown to bind strongly (EC₅₀ 32 ng/mL, 0.33 nM, Fig. 2A) to S309, consistent with the previously reported data (EC₅₀ = 1.5 nM).¹⁵ However, all the synthetic N-glycopeptides, either with or without core-fucose, did not show apparent affinity to S309 (Fig. 2B). To facilitate a more oriented immobilization of the glycopeptides, we also coated the biotinylated glycopeptides for ELISA analysis, but no apparent binding was observed, either (data not shown). Given the fact that the synthetic glycoforms of RBD without core-fucosylation did not show difference in affinity from the non-glycosylated RBD as reported in a recent study,¹⁶ the present data suggest that the core-fucosylated N-glycopeptides alone are not sufficient for high-affinity binding to the S309, and the glycopeptide epitope might need to be present in the context of the RBD domain to maintain an active conformational epitope or, alternatively, additional peptide domains are required as an integral neutralizing epitope for high-affinity binding to S309.

Figure 2.

ELISA of S309 mAb binds to SARS-CoV-2 S protein trimer (3A) and synthetic (glyco)peptides (3B); SPR analysis of the binding between hCLEC4G and SARS-CoV-2 glycopeptides (3C-3I). Biotin tagged glycopeptides were captured on NeutrAvidin immobilized CM5 chips and the hCLEC4G was run as analytes at 2× serial dilutions starting from 2000 nM (final, 3.8 nM). The K_D value was obtained by steady-state fitting.

We next evaluated the binding of human lectin CLEC4G with the synthetic glycopeptides by surface plasmon resonance (SPR) analysis. Thus, the biotinylated glycopeptides (18-24) were immobilized on a Neutravidin coated CM5 chip and lectin CLEC4G was run as the analyte. We found that while the plain peptide (24) (Fig. 2I) and the glycopeptide carrying the Man₅ N-glycan (21) (Fig. 2F) did not bind to hCLEC4G, the glycopeptides (18, 20, 22, and 23) carrying terminal GlcNAc N-glycans at the N343 or N331 site all exhibited substantial binding to hCLEC4G, with a K_D of 1–2 μM (Fig. 2C, E, G and H, respectively). Interestingly, the glycopeptide (23) carrying two GlcNAc-terminated N-glycans at both N331 and N343 sites did not show enhanced affinity to hCLEC4G, demonstrating no clustering effect. Core-fucosylation had little effect on the hCLEC4G binding (18 vs. 20). On the other hand, masking the terminal GlcNAc by galactosylation, as shown in glycopeptide 19, abolished the binding to hCLEC4G (Fig. 2D). These results confirm that lectin hCLEC4G recognizes specifically N-glycans and N-glycopeptides carrying terminal GlcNAc moieties.

In conclusion, a highly efficient chemoenzymatic synthesis of the naturally dominant core fucosylated N-glycopeptides derived from the SARS-CoV-2 spike protein RBD is described. A preliminary study shows that these structurally well-defined, homogeneous N-glycopeptides of SARS-CoV-2 are useful for characterizing the glycan-binding specificity of lectins and potentially the glycan-dependent epitopes of neutralizing antibodies. The chemoenzymatic method described here should be applicable for the synthesis of an expanded library of homogeneous SARS-CoV-2 glycopeptides corresponding to respective domains of the spike protein, which will facilitate the characterization of unique glycopeptide epitopes for vaccine design and therapeutic development.