Glycosylation and S-palmitoylation regulate SARS-CoV-2 spike protein intracellular trafficking

Summary Post-translational modifications (PTMs), such as glycosylation and palmitoylation, are critical to protein folding, stability, intracellular trafficking, and function. Understanding regulation of PTMs of SARS-CoV-2 spike (S) protein could help the therapeutic drug design. Herein, the VSV vector was used to produce SARS-CoV-2 S pseudoviruses to examine the roles of the 611LYQD614 and cysteine-rich motifs in S protein maturation and virus infectivity. Our results show that 611LY612 mutation alters S protein intracellular trafficking and reduces cell surface expression level. It also changes S protein glycosylation pattern and decreases pseudovirus infectivity. The S protein contains four cysteine-rich clusters with clusters I and II as the main palmitoylation sites. Mutations of clusters I and II disrupt S protein trafficking from ER-to-Golgi, suppress pseudovirus production, and reduce spike-mediated membrane fusion activity. Taken together, glycosylation and palmitoylation orchestrate the S protein maturation processing and are critical for S protein-mediated membrane fusion and infection.

The S glycoprotein is a critical target for pathogenic coronavirus vaccine development, and current COVID-19 vaccines employ full-length or portions of S protein as the antigen to induce neutralizing antibodies against SARS-CoV-2 entry (Li et al., 2020;Tregoning et al., 2020). A process by which suboptimal antibodies against viral glycoproteins enhance viral infection through the Fcg receptor (Lee et al., 2020), antibody-dependent enhancement (ADE) has been a concern in vaccine development against pathogens such as dengue virus (Katzelnick et al., 2017;Ulrich et al., 2020), SARS-CoV-1, and MERS-CoV (Wan et al., 2020;Wang et al., 2016). Several studies show that ADE of SARS-CoV-2 is mediated by Fcg receptor IIA or complement component C1q (Maemura et al., 2021;Okuya et al., 2022;Wang et al., 2022). Monoclonal antibodies specific for the 597 LYQD 600 motif of the SARS-CoV-1 S protein are shown to have ADE activity (Wang et al., 2016). An LYQD motif is also present in the SARS-CoV-2 S protein, but whether eliminating the ADE-associated sequence in the S antigen is beneficial for an effective COVID-19 vaccine design remains unknown.
Maturation of the S glycoprotein is critical for coronavirus infection and transmission and can also be one of the antiviral targets. The S protein undergoes several post-translational modifications (PTMs), iScience Article including N-linked glycosylation, palmitoylation, and proteolytic processing as part of its maturation process (Fung and Liu, 2018). 12 out of 23 asparagine residues in the SARS-CoV-1 S protein are glycosylated (Krokhin et al., 2003). SARS-CoV-1 S proteins are glycosylated in the endoplasmic reticulum (ER) with high-mannose glycans, which are then further modified as complex N-glycans in the Golgi (Duan et al., 2020;Nal et al., 2005). Glycosylation can influence viral glycoprotein folding, function, immune evasion, and virus infection (Huang et al., 2021;Watanabe et al., 2019Watanabe et al., , 2020. The endodomains of SARS-CoV-1 and SARS-CoV-2 S proteins contain a cysteine-rich motif for palmitoylation, which might participate in membrane fusion and infectivity (Petit et al., 2007;Wu et al., 2021). It has been shown that zinc finger DHHC domain palmitoyltransferase 5 (zDHHC5) and Golgin subfamily A membrane 7 (GOLGA7) interact with S protein and induce its palmitoylation (Gordon et al., 2020;Wu et al., 2021;Zeng et al., 2021). How these PTMs affect SARS-CoV-2 S protein stability, intracellular trafficking, and function will need to be addressed more thoroughly.
In this study, the vesicular stomatitis virus (VSV) vector was used to generate the SARS-CoV-2 S pseudoviruses for the functional characterization of the LYQD and cysteine-rich motifs on S protein maturation and virus infectivity. Our results suggest that the LYQD motif was involved in the S protein glycosylation process, and the palmitoylation of the cysteine-rich motif participated in the S protein trafficking and maturation process.

RESULTS
Mutations in the LYQD motif change the glycosylation pattern of the SARS-CoV-2 S protein Antibodies recognizing the 587 LYQD 590 and C 593 amino acid residues in the SARS-CoV-1 S protein have ADE activity (Wang et al., 2016). To examine whether the 611 LYQD 614 sequence in SARS-CoV-2 S protein maturation can be removed from the S protein, alanine substitution mutations in the LYQD sequence ( 611 LYAA 614 , 611 AAQD 614 , and 611 AAAA 614 ) were generated in the mammalian expression plasmid (Figure 1A). As the following cysteine residue might be involved in inter-or intra-molecule disulfide bond formation, C 617 of SARS-CoV-2 S protein was not mutated. Wild type (WT) and mutant S protein expression in BHK21 cells was achieved through plasmid DNA transfection, followed by infection with the VSVDG-GFP/G virus to secrete the S pseudotyped virus particles (S pp ) into the culture medium. When the S protein expression in transfected cells was examined by immunoblotting with an anti-S2 antibody, full-length S protein ($180 kDa) and S2 ($100 kDa) were both detected in the cell lysate ( Figure 1B, left panel). However, the S2 proteins of the AAQD and AAAA mutants had different electrophoretic mobility in contrast to those from the WT and LYAA mutants. S pp secreted into the culture supernatant was then examined by immunoblotting with antibodies specific for the S1 and S2 proteins ( Figure 1B, right panel). Very little unprocessed FL S protein was detected in the S pp culture medium. The S1 and S2 proteins from WT and LYAA S pp each appeared as a doublet on an immunoblot, indicating two different protein sizes, but appeared as a single high molecular weight band in the culture medium containing AAQD and AAAA S pp .
Because the SARS-CoV-2 S protein has 22 predicted N-linked glycosylation sites, with the N 603 and N 616 residues located close to the LYQD motif (Watanabe et al., 2020), we postulated that the differential electrophoretic mobility observed for the S1 and S2 mutants could be due to changes in protein glycosylation.
To test this, WT and mutant S pp were treated with peptide N-glycosidase F (PNGase F) to remove N-linked oligosaccharides from the S protein. The protein size of S1 and S2 proteins from WT, AAQD, and AAAA S pp became the same after PNGase F treatment ( Figure 1C), suggesting that the 611 LY 612 mutation altered the glycosylation pattern of S protein. When the S pp titer was measured by assessing the GFP signal in infected BHK21-hACE2 cells, the AAQD and AAAA S pp had significantly lower viral titer compared to the WT and LYAA S pp ( Figure 1D). Taken together, these data suggest that mutation of LY residues affects the glycosylation pattern of S protein. To test whether the LYQD sequence is also required for S pp production in other cell types, HEK293T was used for S pp packaging ( Figures 1E and 1F). While the WT and LYAA mutant proteins show normal S protein processing and S pp production, only one dominant S1 and S2 form were present in S pp from HEK293T (Figure 1E), which was different from the S pp generated from BHK-21 cells. The S2 subunit of the AAQD and AAAA mutants might be unstable and could not be detected in HEK293T cell and medium. Surprisingly, a large amount of S1 subunit was detected in the culture medium. When the virus particles were precipitated with PEG ( Figure S1), S1 and S2 subunits were present in the WT and LYAA S pp but not in the AAQD and AAAA S pp. This suggests that the S1 subunit in AAQD and AAAA mutants was present in the culture medium but not associated with virus particles. Viral titers of AAQD and AAAA S pp were also consistently reduced ( Figure 1F).
The SARS-CoV-2 S protein contains 22 asparagine (N) residues, 16 of which have N-linked glycosylation (Shajahan et al., 2020;Walls et al., 2020;Watanabe et al., 2020). The N 603 and N 616 residues are close to 611 LYQD 614 sequence. Whether N 603 and N 616 mutation to alanine in S protein might show a similar phenotype as LY mutation was further tested. As shown in Figure S2A, the protein expression level and glycosylation pattern of the N603A and N616A mutants were very similar to the WT protein, suggesting that the phenotype of the 611 LY 612 mutation is not derived from the dysregulation of the neighboring glycosylation sites. Interestingly, the infectivity of the N603A and N616A S pp was significantly reduced ( Figure S2B), indicating that appropriate glycosylation is critical for S protein-mediated infection.
The C-terminal tail of the S protein contains an ER-retention motif, and deletion of the tail facilitates S protein targeting to the plasma membrane for pseudovirus production (Lontok et al., 2004;McBride et al., 2007;Xiong et al., 2020). The S mutants with truncation of the C-terminal 19 amino acids (SD19) and LYQD-related mutations were generated to evaluate S pp production efficiency. The WT and LYAA SD19 pp were produced with high efficiency, as indicated by immunoblotting and virus titration ( Figure S3). The AAQD and AAAA SD19 pp production were much lower than the WT and LYAA SD19 pp . These data indicate that the LY sequence was required for a high titer of S pp and SD19 pp production.

LY 612 sequence is required for S protein maturation and trafficking
Protein glycosylation with diverse glycan occurs in the ER and Golgi. As 611 LY 612 mutation altered the glycosylation profile on both S1 and S2 subunits, we next examined whether 611 LY 612 sequence affects S protein subcellular distribution. When the expression pattern of S protein was examined by immunostaining with antibodies recognizing the S2 epitope (1A9), WT, LYAA, and AAQD S proteins were present in the cytoplasm and partially colocalized with the ER marker, which might correlate with the nascent S protein localization ( Figure 2A). When an anti-S2 antibody (ECD45) with a high affinity to the prefusion state of trimeric S (mature S) was used for immunostaining, the mature WT and LYAA S protein expression were concentrated in the Golgi, and some speckles in the cytoplasm ( Figure 2B). LY mutation showed a slightly reduced mature S expression compared to WT by immunostaining. In contrast, the signals for mature S of the AAQD mutant were very weak, and rarely present in the Golgi, suggesting that the LY mutation downregulates the S protein trafficking and maturation process. Some S proteins might further transport from Golgi to the cell surface, which may contribute to cell-cell fusion and spreading of SARS-CoV-2. Hence, whether the LYQD mutation affects S protein expression on the cell surface with anti-S2 antibodies (1A9 and ECD45) was examined by flow cytometry. As shown in Figure 2C, a higher percentage of cells had mature S protein expression on the cell surface in the WT and LYAA plasmid-transfected cells than in the AAQD-transfected group. The percentages of intracellular S protein expression of the WT, LYAA, and AAQD plasmid-transfected cells did not show any significant difference. The mean fluorescence intensity (MFI) of the LYAA mutant on the cell surface or intracellular was lower compared to the WT, suggesting that the protein stability of the LYAA mutant might also be slightly affected. Taken together, the LYQD sequence might sustain S protein stability and facilitate S protein moving from ER to Golgi for glycan modification and further to the plasma membrane during its maturation.

Palmitoylation is required for efficient S pp and SARS-CoV-2 production
Protein palmitoylation primarily functions in protein subcellular trafficking. The endodomain of SARS-CoV-2 S protein contains a cysteine-rich motif for palmitoylation. To evaluate the function of palmitoylation in S protein maturation and virus infectivity, full-length S protein expression and S pp packaging in HEK293T cells were examined in the presence of a general palmitoylation inhibitor, 2-bromopalmitate (2BP iScience Article S, S1, and S2 protein intracellular expression was not altered by the 2BP treatment at non-toxic concentrations (1-10 mM; Figures 3A and S4), the S pp packaging and titer in the culture medium were reduced ( Figures 3A and 3B). As the cysteine-rich motif remains intact in the SD19 protein, the SD19 pp production was also inhibited by the 2BP treatment ( Figure S5). Genetic lineages of SARS-CoV-2 continue to evolve some variants, such as alpha, beta, gamma, and delta variants, which may have a higher transmission   Figures 3C and 3D) but did not affect the S, S1, and S2 intracellular expression ( Figure S6). To examine the importance of S protein palmitoylation in SARS-CoV-2 virus replication, Vero E6 cells were infected with SARS-CoV-2 virus (moi = 1) in the presence of 2BP treatment for 48 h. As shown in Figure 3E, palmitoylation inhibition did not affect intracellular S protein expression but blocked SARS-CoV-2 virus production in the culture medium. The virus titer was slightly lower in the presence of 2BP ( Figure 3F). In summary, palmitoylation inhibition suppresses S pp and SARS-CoV-2 virus production.

SARS-CoV-2 S protein is palmitoylated at multiple cysteine residues
To confirm S protein palmitoylation was inhibited by 2BP, the palmitoylated S protein was monitored by using a modified protocol of the acyl-PEG exchange gel . As shown in Figure 4A, there were at least three palmitoylated species of the S protein as observed by the slower migrating bands on the Western blot after acyl-PEG exchange. The palmitoylation level of S protein was inhibited by the 2BP treatment. We also detected a single major palmitoylated species of the S2 protein with or without 2BP treatment. Similarly, the palmitoylation level of SD19 was also suppressed by the 2BP treatment ( Figure S7). Furthermore, the palmitoylated S2 protein was also incorporated in the S pp virus ( Figure 4B), suggesting that S protein palmitoylation might directly participate in virus packaging.
The cysteine-rich motifs of SARS-CoV-1 and SARS-CoV-2 S protein each contain four clusters known to be involved in protein palmitoylation. To examine the functional importance of the cysteine-rich clusters in SARS-CoV-2 S protein palmitoylation and S pp production, we generated cysteine-to-serine mutations in the cysteine-rich motif of S protein (SS-I, SSS-II, SSS-III, SS-IV, SS-I + II, and SS-III + IV) ( Figure 4C). Consistent with previous reports (Petit et al., 2007;Wu et al., 2021), clusters I and II are crucial for SARS-CoV-2 S protein palmitoylation. SS-I and SSS-II mutants showed drastically diminished palmitoylation levels and palmitoylated forms S protein were barely detectable for the SS-I + II mutant ( Figure 4D). In contrast, the major palmitoylated species of the SARS-CoV-2 S protein were still observed for the SSS-III, SS-IV, and SS-III + IV mutants. S protein palmitoylation is important for S pp production because S2 expression level in S pp and S pp titers was reduced in the SS-I, SSS-II, and SS-I + II mutants ( Figures 4E, 4F, and S8). Interestingly, the SSS-III, SS-IV, and SS-III + IV mutants showed higher S pp production but reduced virus titers ( Figures 4E and S8), suggesting that cysteine clusters III and IV might have additional roles in S protein processing and S pp production that are independent of protein palmitoylation. Collectively, these results demonstrate that cysteine clusters I and II are important for S protein palmitoylation and S pp production.

Palmitoylation targets mature S protein to the Golgi and plasma membrane
As palmitoylation may influence protein localization, trafficking, and stability (Aicart-Ramos et al., 2011; Yang et al., 2020), we evaluated whether palmitoylation modulates S protein subcellular localization. The localization of nascent S protein was unaffected with 2BP treatment when the S protein was stained with anti-S1 antibody (clone HL263) ( Figure 5A). In contrast, the localization of mature S protein stained by ECD45 antibody was concentrated in the Golgi, plasma membrane, and speckles in the cytoplasm. Upon 2BP treatment, the fluorescent intensity of mature S protein was reduced and dispersed as an ERlike distribution ( Figure 5B). The S protein subcellular localization of the cysteine cluster mutants was also examined. Similarly, consistent with palmitoylation targeting mature S protein to the Golgi, the palmitoylation-defective SS-I and SSS-II mutant S proteins were distributed outside the Golgi ( Figure 5C). On the other hand, the mature SSS-III and SS-IV S proteins remained predominantly concentrated in the Golgi and plasma membrane. Furthermore, the mature SS-I + II mutant showed a more dispersed compartment outside the Golgi, and the SS-III + IV combination mutant was even more condensed in the Golgi and plasma membrane despite similar localization of the nascent S proteins ( Figure S9). In summary, palmitoylation is important for the subcellular localization and trafficking of the S protein.  iScience Article palmitoylated S proteins observed at the plasma membrane ( Figure 5), we examined whether palmitoylation is required for S protein-mediated cell fusion. GFP-and S-expressing BHK-21 cells were overlaid on the calu-3 cell film at 4 C for 45 min. After washing with PBS, the S protein-ACE2-mediated cell fusion between BHK-21 and Calu3 was monitored by tracing the GFP + multinucleated cells after 4 h incubation ( Figure 6A). The GFP + cells without S protein expression remained as single cells (300-400 mm 2 ) while those with S protein expression became large multinucleate cells (1500-2000 mm 2 ) upon cell fusion ( Figures 6B and 6C). Upon palmitoylation inhibition by 2BP, the GFP + multinucleate cells had reduced cell-cell fusion (1000 mm 2 on average). Consistent with S palmitoylation being important for spike-mediated membrane fusion, significantly reduced cell fusion was observed for palmitoylation-defective SS-I and SSS-II mutants ( Figures 6D and 6E). Further highlighting the different functions of clusters III and IV, cell-cell fusion was not reduced when these clusters were mutated. Interestingly, SS-IV mutation enhanced the cell-cell fusion by about 50%, although likely through palmitoylation-independent mechanisms. Taken together, these data iScience Article indicate that the palmitoylation of S protein in cysteine clusters I and II is involved in the mature S protein trafficking to the Golgi and cell surface to facilitate S-ACE2-mediated cell fusion.

DISCUSSION
The VSVDG-based pseudovirus was applied to characterize SARS-CoV-2 S protein maturation and virus infection. The 611 LY 612 sequence in the putative ADE motif of S protein is required for correct glycosylation and maturation, which is essential for the production of infectious S pp particles. In contrast, the S protein can tolerate 613 QD 614 mutation in the putative ADE motif without affecting glycosylation and virus infection. The 613 QD 614 could be further evaluated in a vaccination setting. Palmitoylation targets the mature S protein trafficking to the Golgi and plasma membrane, which subsequently facilitates pseudovirus packaging and cell-cell membrane fusion. The four cysteine clusters of S protein have distinct roles in these processes. The study reveals that glycosylation and palmitoylation coordinate the S protein maturation processes, which are critical for the protein to function. (B and C) The S and GFP coexpression plasmids were introduced to BHK21 cells by DNA transfection for 24 h. The transfected cells were collected and overlaid on Calu-3 cells at 4 C for 1 h, and the unbound cells were removed with PBS wash. After further 4 h incubation, the images of the GFP + cells were captured by fluorescent microscope. GFP + cell sizes in five randomly selected fields were quantified using ImageJ (C).
(D and E) The cysteine mutants were subjected to the cell fusion assays (D). GFP + cell sizes in five randomly selected fields were quantified using ImageJ (E). Scale bars: 100 mm **p < 0.01. Error bars represent SEM and n = 3. iScience Article Glycosylation on the newly translated protein facilitates correct protein folding. Glycans could influence the conformation of S protein (Sztain et al., 2021), affect the binding affinity between S protein and ACE2 receptor (Casalino et al., 2020;Koehler et al., 2020), and avoid S protein from antibody recognition (Wang, 2020;Watanabe et al., 2020;Zhao et al., 2021). Based on our results, the S protein in S pp showed two different glycosylation forms in BHK21 cells but only one form (lower molecular weight) in 293T cells. The difference could be due to the diversity of glycosylation machinery in various cell types or fast degradation of the S protein with alternative glycosylation in 293T cells. The 611 LY 612 S protein with an altered glycosylation profile (higher molecular weight) could be incorporated in S pp in BHK21 cells, and the 611 LY 612 mutant S pp showed lower infectivity. The SARS-CoV-2 S protein contains multiple N-linked glycosylation sites, and N 603 & N 616 residues are close to 611 LYQD 614 sequence. The glycosylation levels of the N603A and N616A mutants are different from the 611 LY 612 mutant, suggesting that the low maturation process of the 611 LY 612 mutant is not due to the changes in the glycosylation of the neighboring amino acid sequence. Although the LY sequence is located in the S1 subunit, the S2 glycosylation profile was also affected, suggesting that the LY mutation keeps the S protein from trafficking to the Golgi, where glycan modification on both S1 and S2 subunits can be further processed. In 293T/17 cells, the S1 and S2 subunits of 611 LY 612 mutant might be unstable and subjected to protein degradation. It is also possible that the 611 LY 612 might impact protein folding, leading to ER quality control retention (Benyair et al., 2011).
Palmitoylation is important for SARS-CoV-1 and SARS-CoV-2 S protein maturation in order to function during virus infection and membrane fusion. Palmitoylation might facilitate S protein trafficking from ER to the cell surface during its maturation. The cysteine cluster I and II of SARS-CoV-1 and SARS-CoV-2 S proteins are the main palmitoylation site, as shown in our study and several studies (Mesquita et al., 2021;Petit et al., 2007;Puthenveetil et al., 2021). Host proteins zDHHC5 and GOLGA7 are involved in the palmitoylation of S protein (Gordon et al., 2020;Wu et al., 2021;Zeng et al., 2021). zDHHC enzymes such as zDHHC2, zDHHC3, zDHHC8, zDHHC9, and zDHHC20 may also contribute to the palmitoylation process of S protein (Mesquita et al., 2021;Puthenveetil et al., 2021;Ramadan et al., 2022). zDHHC enzymes distribute at ER, Golgi, and plasma membranes throughout the cellular secretory pathway (Malgapo and Linder, 2021). Zeng et al. showed that the intracellular distribution of S protein was not altered in the zDHHC5 or GOLGA7 knockout condition (Zeng et al., 2021), but our data show that mutations of the palmitoylation sites reduce the Golgi localization of the mature S protein. The discrepancy may be due to the difference in antibody selection for immunostaining. The other possibility was that the zDHHC family enzymes might compensate for the deficiency of the zDHHC5 or GOLGA7 and complete the palmitoylation of S protein. Mutations in cysteine clusters I and II also lead to a deficient ER-to-cell surface trafficking of the S protein, S1/S2 cleavage, virus packaging, and S-mediated membrane fusion. The cluster I and II mutants carry two palmitoylation sites of S proteins, whereas cluster III and IV mutants might contain more than two palmitoylation sites of S proteins. Moreover, the mutations in the cysteine clusters III and IV seem further enhance Golgi localization, S1/S2 cleavage, and virus packaging, suggesting that they might have a regulatory role in these processes. Based on the S protein structure prediction (Zheng et al., 2021), the cysteine-rich motif is located in the cytoplasmic tail proximally to the transmembrane domain. The cysteine-rich clusters I and II are in a stable alpha-helix structure and might be more accessible for zDHHC enzyme, which is known to interact with SARS-CoV-2 S protein. As the cysteine-rich motif in SARS-CoV-2 S is highly conserved in all the variants, S palmitoylation might be a potential target for antiviral drug development.
S protein trafficking is via the secretory pathway (Bracquemond and Muriaux, 2021;Mendonca et al., 2020). After synthesized, S protein is modified in ER and Golgi for post-translational modifications. It has been shown that the S protein has ER export and ER retrieval signals at the cytoplasmic tail (Cattin-Ortola et al., 2021). How the LYQD motif-associated glycosylation regulation and palmitoylation of S protein coordinate with these ER export and ER retrieval signals in the maturation process will be worth further investigation.

Limitations of the study
First, we have utilized vesicular stomatitis virus pseudoviruses to study glycosylation and palmitoylation of S protein. However, it is difficult to develop SARS-CoV-2 infectious cDNA clone to understand glycosylation and palmitoylation during virus replication and package. Second, detecting ADE activity from AAQD and AAAA S pp is not very easy because viral titers of AAQD and AAAA S pp were lower than that of LYAA S pp . Further studies will be required to evaluate the glycosylation mechanisms of LYQD motif and palmitoylation of S protein coordinate with these ER signal peptides in the maturation process.

ACKNOWLEDGMENTS
The SARS-CoV-2-related experiments were performed by the P3 research team at NHRI. The confocal images were obtained with assistance from the optical biology core facility of NHRI. The ECD45 antibody was kindly provided by AnTaimmu BioMed. The study was supported by the Ministry of Health and Welfare, Taiwan

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Guann-Yi Yu (guannyiy@nhri.edu.tw).

Materials availability
Plasmids in this paper will be shared by the lead contact upon request.
Data and code availability d Data reported in this paper will be shared by the lead contact upon request.
d There is no original code associated with this work.
d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Cell, virus, and reagent
Immunofluorescence assay 293T/17 cells were transfected with plasmid DNA to express full-length or mutant S protein on coverslips, and the cells were fixed in 4% paraformaldehyde and permeabilized in 0.1% Triton X-100 of PBS buffer at ll OPEN ACCESS