key: cord-0954182-lop16nqn
authors: Huang, Han-Yi; Liao, Hsin-Yu; Chen, Xiaorui; Wang, Szu-Wen; Cheng, Cheng-Wei; Shahed-Al-Mahmud, Md.; Chen, Ting-Hua; Lo, Jennifer M.; Liu, Yo-Min; Wu, Yi-Min; Ma, Hsiu-Hua; Chang, Yi-Hsuan; Tsai, Ho-Yang; Chou, Yu-Chi; Hsueh, Yi-Ping; Tsai, Ching-Yen; Huang, Pau-Yi; Chang, Sui-Yuan; Chao, Tai-Ling; Kao, Han-Chieh; Tsai, Ya-Min; Chen, Yen-Hui; Wu, Chung-Yi; Jan, Jia-Tsrong; Cheng, Ting-Jen Rachel; Lin, Kuo-I; Ma, Che; Wong, Chi-Huey
title: Impact of glycosylation on a broad-spectrum vaccine against SARS-CoV-2
date: 2021-08-26
journal: bioRxiv
DOI: 10.1101/2021.05.25.445523
sha: b4864af4529e6d9920b2f7f1e1e142b113498178
doc_id: 954182
cord_uid: lop16nqn

A major challenge to end the pandemic caused by SARS-CoV-2 is to develop a broadly protective vaccine. As the key immunogen, the spike protein is frequently mutated with conserved epitopes shielded by glycans. Here, we reveal that spike glycosylation has site-differential effects on viral infectivity and lung epithelial cells generate spike with more infective glycoforms. Compared to the fully glycosylated spike, immunization of spike protein with N-glycans trimmed to the monoglycosylated state (Smg) elicits stronger immune responses and better protection for hACE2 transgenic mice against variants of concern. In addition, a broadly neutralizing monoclonal antibody was identified from the Smg immunized mice, demonstrating that removal of glycan shields to better expose the conserved sequences is an effective and simple approach to broad-spectrum vaccine development. One-Sentence Summary Removing glycan shields to expose conserved epitopes is an effective approach to develop a broad-spectrum SARS-CoV-2 vaccine.

SARS-CoV-2 spike (S) protein contains 22 N-linked and at least two O-linked glycosylation sites per monomer, which are important for S protein folding and processing, and for evading immune response by shielding specific epitopes from antibody neutralization (1, 2). The glycan profiles of S protein expressed from various cell sources have been reported, revealing a conserved pattern of 8 specific sites harboring >30% underprocessed high-mannose- 5 type and hybrid-type N-glycans, with the rest 14 sites predominantly in the complex-type (1, 3) . It was shown that a trimeric S protein with complex-type glycoform is more efficient in receptor recognition and viral entry than high-mannose variants derived either from GnTI -/-HEK293S (4) or MGAT1 -/-HEK293T (5) . We investigated the glycosylation of S protein expressed from lung epithelial cells, the primary cells for infection, and found that complex-type glycans and 10 sialylation of S protein are required for higher avidity to the receptor ( fig. S1A-C) . We further explored the S protein glycosylation effect on viral infectivity and its correlation with sequence and glycoform conservation, and provides insights into the design of a broad-spectrum vaccine. A full panel of 24 lentivirus-based pseudovirus variants (22 N-and 2 O-glycosites) were generated for evaluating the viral entry efficiency in five hACE2-expressing cell lines, including 15 HEK293T, Vero-E6, and three human lung cell lines, A549, Calu-1 and Calu-3 cells (Fig. 1B,  C) . A significant reduction of infectivity was observed for two mutations in RBD, N331Q and N343Q, consistent with the previous report (6) , and mutations of the two O-glycosites (T323A and S325A), despite their low occupancy (Fig. 1A, C) (1). Similarly, deletion of N122 glycosylation in NTD resulted in low protein expression and reduced infectivity (Fig. 1C, fig.  20 S1F). However, the two NTD mutations, N149Q and N165Q, increased infectivity in Vero-E6 and Calu-1 cells, respectively (Fig. 1C) . Most importantly, we identified two mutants, N801Q and N1194Q (Fig. 1C ) that universally abolished virus infectivity in all five cells. These mutations caused low-yield expression ( fig. S1F) , and the N801Q mutant was more prone to degradation while the N1194Q mutant disrupted S protein trimerization ( fig. S1G, H) . 25 The glycan profile analysis of S protein revealed a higher abundance of complex-type (76%) and much fewer hybrid-type glycans (1%) from the human lung epithelial cell line BEAS-2B compared to the human kidney epithelial cell line HEK293T (61% and 23%, respectively) ( Fig. 1D-E, fig. S2, 3, fig. S4A, B) . Among the N-glycosites, Man5 is the predominant oligomannose-type glycan found across the HEK293T-expressed S protein, while it is only seen 30 at site N61 from BEAS-2B ( fig. S2, 3 ). In addition, the complex-type glycans at N74, N149, N282 and N1194 in BEAS-2B are more diversely processed (multiple antennae, galactosylation, fucosylation or sialylation) than in HEK293T cells, while those at N122, N331, N1098 and N1134 are less ( fig. S2, 3) , and N149 and N17 harbor no core fucose in BEAS-2B cells. More importantly, we observed an overall higher degree of sialylation on all 22 N-glycosites from 35 BEAS-2B (51%) than from HEK293T (35%), HEK293E (26%) ( fig. S4 ) or HEK293F cells (15%) (1). Particularly, the two N-glycosites (N331 and N343) of RBD are more sialylated in BEAS-2B (98% and 59%) than in HEK293T (49% and 15%) ( fig. S4A, B) . Despite the differences, the S protein from all cell types contains a non-complex-type glycan belt located at the middle stem region of S2 (Fig. 1F, fig. S4D ), in which there is a critical site for infection 40 (N801) (Fig. 1C) , a diverse-glycoform site N1074 ( fig. S2, 3) and a site essential for S protein expression (N717) ( fig. S1F, I) .

From the S protein glycosylation profile, we conducted structural analysis of glycan coverage over protein surface areas and overlaid with multiple alignment results using 1,117,474 SARS-CoV-2 S protein sequences extracted from GISAID (7) (version: Apr. 18, 2021) . It was 45

shown that most of the glycosylation sites are highly conserved ( fig. S5A) , and the RBD, Fusion Peptide (FP), and Heptad Repeat 2 (HR2) domains contain higher percentage of conserved surface residues (33.6%, 42.1% and 45.7%) ( fig. S5B, C) . Incorporating these data into S protein 3D structure mapping revealed several conserved surface regions that are shielded by glycans, including the lower flank of RBD, the non-complex-type glycan belt and HR2 (Fig. 1G, H, fig.  S5D ), leading to the thought that exposing the glycan-shielded conserved regions may elicit immune response in a more efficient way. 5 Our initial attempt to mutate multiple glycosites led to a dramatically reduced expression of S protein ( fig. S1F ). This problem however may not be encountered with the use of mRNA as vaccine, as when the glycoengineered RNA is inside the antigen presenting cells, the translated and unfolded S protein mutant may still be processed for antigen presentation. Here we expressed S protein from GnTI -/cells to produce high-mannose glycoforms and trimmed the 10 glycans with Endoglycosidase H (Endo H) to N-acetylglucosamine (GlcNAc) at each Nglycosite ( fig. S6A , B) (8, 9) , generating the monoglycosylated S protein (Smg) ( Fig. 2A, fig.  S6C -F). This modified Smg and the fully glycosylated S protein (Sfg), both with essentially the same trimeric structure in solution ( fig. S6G, H) , were used to immunize BALB/c mice (n=5) with two intramuscular injections (Fig. 2B) . Sfg is similar to the immunogens used in many 15 current COVID-19 vaccines which were either approved or in clinical trials, including the insectcell expressed S protein vaccines from Sanofi and Novavax (10), the CHO cell expressed recombinant S vaccine from Medigen (11), the adenovirus-based vaccines from AstraZeneca and J&J, and mRNA vaccines from Pfizer-BioNTech and Moderna (10) .

Mice immunized with Smg induced superior humoral immune response after second 20

immunization as compared to Sfg, with 2-fold higher IgG titer against S protein (Fig. 2B , C) and stronger antibody neutralization based on the inhibition of SARS-CoV-2 cytopathic effect (CPE) (Fig. 2D) . The analysis of IgG subtype titer and IFN-g and IL-4 production by Tfh cells revealed that Smg vaccine induced more balanced Th1/Th2 response while Sfg vaccine elicited Th2-biased responses which may cause vaccine-associated enhanced respiratory disease (VAERD) (Fig. 2E-25 G, fig. S7A -C) (10) . Furthermore, Smg vaccine induces higher frequency of IL-21 + Tfh cells (Fig. 2H ) as well as the elevated levels of granzyme B-producing CD8 T cells (Fig. 2I) , indicating that more potent humoral and cellular adaptive immunity was elicited by Smg, as compared with that induced by Sfg. We then examined the frequency of S protein-specific B cells (CD3 − CD19 + S + ) from the spleen of mice immunized with Sfg or Smg, and found that mice 30

immunized with Smg generated more S protein-specific B cells (Fig. 2J, fig. S9B ). The B cell repertoire analysis from Sfg and Smg immunized mice indicated that in the Smg group, several loci of IGHV (Fig. 2K, fig. S7F ) and IGKV (Fig. 2L, fig. S7G ) were more presented and more lambda light chain was used (Sfg, 1.92%; Smg, 9.68%) (Fig. 2M ). In addition, Smg vaccination also exhibited a broad-spectrum protection against SARS-CoV-2 variants including the D614G, 35

Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2) (12) and several other human coronaviruses with better IgG titers (Fig. 2N, fig. S7H ) and enhanced neutralizing antibody responses ( Fig. 2O , P) as compared to Sfg. These variants are of particular concern nowadays as they showed resistance to several therapeutic antibodies and convalescent sera, and exhibited reduced protection by many approved vaccines (13) (14) (15) (16) (17) . 40

To evaluate the in vivo protective efficacy of Smg vaccine against SARS-CoV-2, we then first carried out wildtype SARS-CoV-2 (hCoV-19/Taiwan/4/2020) challenge in Syrian hamsters (Fig.  3A) , and Smg vaccinated hamsters (n = 5) showed less reduction in body weight as compared to the Sfg and PBS groups (Fig. 3B) and similar virus titer reductions in the lungs of both Sfg and Smg vaccinated hamsters (Fig. 3C) . Since hamsters only showed mild-to-moderate sickness upon 45 SARS-CoV-2 infection, we also used the severe-disease model of highly susceptible hACE2 transgenic mice (n = 3) challenged with wildtype SARS-CoV-2 intranasally (Fig. 3A) . The analysis of IgG neutralizing titer and IgG subtypes against S protein showed similar results to the BALB/c mice (Fig 3D-F, fig. S7D-E) but the Sfg group failed to protect mice from severe disease (0% survival rate at 7 dpi), with a rapid decease of body weights (up to 20% loss at 7 dpi for all) (Fig. 3G) . In contrast, the Smg group exhibited 100% survival rate with only mild weight loss (~ 5.75% at 7 dpi) (Fig. 3G, H) . Also, according to the histopathological staining and 5

immunostaining, less lesion was found and almost no virus was detected in the lungs ofimmunized hamsters and hACE2 mice compared to control group (Fig. 3I, J, fig. S8 ). We then evaluated the Smg efficacy against Alpha variant (n=5) and found that Smg vaccination provided better protection with significant less weight reduction and 100% survival rate until 14 dpi (Fig.  3K, L) . The Smg vaccinated mice also showed 75% survival rate in Delta variant challenge (n=4) 10 while only 25% of Sfg vaccinated mice survived until 14 dpi. The significant improved in vivo protection by Smg vaccine provides further evidence that removal of glycan shield from an immunogen is an advantageous strategy to elicit superior immune response.

The sorting of S protein-specific B cells from Smg immunized mice led to the identification of a monoclonal antibody (mAb) m31A7 from IGHV1-18 amplified clones (Fig. 2I, fig. S7F ). 15 This mAb interacts with the full-length S protein, S1 and RBD, but not S2 (Fig. 4A ) and binds to HEK293T cells that express the S protein from different SARS-CoV-2 variants (Fig. 4B ). In addition, m31A7 was shown to neutralize various pseudovirus variants (WT, D614G, Alpha, Beta, and Delta) at sub-picomolar IC50 which is up to 1000-fold higher than the reported human mAb EY6A (18) (Fig. 4C and fig. S9C) . A prophylactic study also demonstrated good in vivo 20 efficacy of m31A7 in hACE2 mice (Fig. 4D) in maintaining both body weight and temperature (Fig. 4E, F) . Bio-layer interferometry (BLI) analysis measured the dissociation constant of m31A7 and its Fab binding to S protein at 34.9pM and 0.22nM, respectively (Fig. 4G) , and epitope mapping by hydrogen-deuterium exchange mass spectrometry (HDX-MS) showed the binding regions on RBD (Fig. 4H, fig. S10A ), which overlap with or locate nearby the observed 25 epitope in cryo-EM structure of S protein in complex with m31A7-Fab (Fig. 4I, L, fig. S10B ).

The structure further revealed a 3-RBD-up conformation with the N165-glycan from neighboring NTD in close vicinity of RBD-m31A7 interface (Fig. 4J) , and the footprint of m31A7 on RBD is similar to human VH1-58 class (Fig. 4K) bypassing most of key mutated residues of variants of concerns (VOCs) (Fig. 4L ) (13, 19, 20) ; yet approaching RBD from a different angle and 30 extending its coverage toward the left flank of RBD (Fig. 4L ) (26), which otherwise could be shielded by N165-glycan. The extremely low usage of IGHV1-18 in the Sfg B cell repertoire (Fig. 2I, fig. S7F ) suggested that Sfg may not elicit m31A7 or related antibodies.

In conclusion, SARS-CoV-2 S protein glycosylation has major influence on virus infection, protein integrity and immune response. The S protein from lung epithelial cells contains more 35 sialylated complex-type glycans to facilitate receptor binding, and glycosites N801 and N1194 are essential for S protein folding and viral infection. The analysis of cell-specific glycoform distribution, sequence conservation and glycan shielding, and their mutual correlations has led to the design of Smg vaccine, in which essentially all glycan shields are removed, making the highly conserved epitopes better exposed to the immune system so that more effective and broadly 40 protective B cell and T cell responses against the virus and variants are elicited (21). As illustrated by the broadly neutralizing antibody shown here and more to be identified in the future, the conserved epitopes targeted by such antibodies could be used for next-generation vaccine development. In an effort to develop vaccines against coronavirus and variants (22) (23) (24) (25) , 

Figs. S1 to S10

References (26-49) The Kappa and Lambda light chain usage. (N) Anti-S IgG titer of sera (shown as fold of increase) with S proteins from SARS-CoV-2 variants and other human coronaviruses (sequence identity shown as %). (O and P) Neutralization potency of sera against pseudotyped and realvirus SARS-CoV-2 variants. pNT50 and PRNT50 represent the reciprocal dilution achieving 50% neutralization. *P < 0.05; **P <0.01. 5

Immunization schedule for Syrian hamster or hACE2 transgenic mice. Sfg (blue), Smg (red) and control (grey). Comparisons are performed by Student's t-test (unpaired, two tailed). * without*P < 0.05; **P <0.01. 7BEN, 7K4N and 7MLZ) onto m31A7-bound RBD. RBD subdomain categorization follows a previous report (20) . (L) Footprints of COV253 (pink) and m31A7 (green) on RBD (grey), with residues of VOCs drawn as red spheres. The additional coverage area by m31A7 is 15 highlighted by a star. *P < 0.05; **P <0.01; ***P < 0.001.

Site-specific glycan 5 analysis of the SARS-CoV-2 spike

Virus-Receptor Interactions of Glycosylated SARS-CoV-2 Spike and Human ACE2 Receptor

Site-Specific Steric Control of SARS-CoV-2 Spike Glycosylation

Multimerization-and glycosylation-dependent receptor binding of SARS-CoV-2 spike proteins

Inhibition of SARS-CoV-2 viral entry upon blocking N-and O-glycan elaboration

The Impact of Mutations in SARS-CoV-2 Spike on Viral Infectivity and 15 Antigenicity

Buckland-Merrett, Data, disease and diplomacy: GISAID's innovative contribution to global health

Vaccination of monoglycosylated hemagglutinin induces cross-strain protection against influenza virus infections

Glycans on influenza hemagglutinin affect receptor binding and immune response

Immunogenicity of clinically relevant SARS-CoV-2 vaccines in nonhuman primates and humans

Development of CpG-adjuvanted stable prefusion SARS-CoV-2 spike antigen as a subunit vaccine against COVID-19

SARS-CoV-2 Variants and Vaccines

SARS-CoV-2 variants, spike mutations and immune escape

Multiple SARS-CoV-2 variants escape neutralization by vaccine-induced humoral immunity

Susceptibility of Circulating SARS-CoV-2 Variants to Neutralization

Effectiveness of Covid-19 Vaccines against the B.1.617.2 (Delta) Variant

Tackling COVID-19 with neutralizing monoclonal antibodies

Structural basis for the neutralization of SARS-CoV-2 by an antibody 40 from a convalescent patient

Ultrapotent antibodies against diverse and highly transmissible SARS-CoV-2 variants

The antigenic anatomy of SARS-CoV-2 receptor binding domain

Immunological mechanisms of vaccineinduced protection against COVID-19 in humans

Chimeric spike mRNA vaccines protect against Sarbecovirus challenge in mice

Mosaic nanoparticles elicit cross-reactive immune responses to zoonotic coronaviruses in mice

Neutralizing antibody vaccine for pandemic and pre-emergent coronaviruses

A Single Immunization with Spike-Functionalized Ferritin Vaccines Elicits Neutralizing Antibody Responses against SARS-CoV-2 in Mice