Furin cleavage of the SARS-CoV-2 spike is modulated by O -glycosylation

1The novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), which is responsible for the current coronavirus disease 2019 (COVID-19) pandemic, contains a unique insertion of 4 amino acids (PRRA) at the S1/S2 boundary of the spike protein (S) that is not present in SARS-CoV and other related coronaviruses (Fig. 1A). This insertion generates a furin cleavage site that has been shown to increase pseudoviral infectivity and syncytia formation in cell culture (Fig. 1A). Recent structural studies have shown the presence of both prefusion (containing S1 and S2) and postfusion (containing S2 only) S structures on virions independent of target cells, indicating that the furin site is being used as virions are made in host cells. The authors suggest that the furin-cleaved S2 structures may play a role in immune system evasion, further highlighting the importance of furin cleavage and modifying factors in viral infectivity/tropism. The region around the furin cleavage site contains predicted sites of O-glycosylation, two of which have recently been confirmed in cell culture (Fig. 1A). O-glycosylation is known to both positively and negatively influence proteolytic cleavage events in diverse proteins, suggesting the possibility that furin cleavage of S may be modulated by O-glycosylation.

2To investigate the role of O-glycosylation in modulating furin cleavage of S, we expressed GFP- or V5-labeled full-length S with or without the novel furin cleavage site (Fig. 1B) in a cell culture background where no O-glycosylation was detected and furin cleavage was active. As shown in Fig. 1C, S expressed in Drosophila S2R+ cells results in full-length S as well as S1 and S2 cleavage products that are the result of furin cleavage; mutation of the furin site (RRAR to AAAR; SPΔfurin-V5) abrogated cleavage, resulting in the presence of only intact S (Fig. 1D). Likewise, treatment with furin inhibitors decreased cleavage (Fig. S1A). S expressed in these cells displayed no reactivity to the O-glycan-specific lectins Helix pomatia (HPA) or Peanut agglutinin (PNA) (Fig. 1E and Fig. S1B), confirming no detectable background O-glycosylation. Additionally, S expressed in these cells retained functional binding ability, as evidenced by binding to cells expressing the receptor hACE2 (Fig. 1F–G). We therefore used this cell background to screen for members of the human UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferase (GALNT) enzyme family that are capable of glycosylating S. Interestingly, only certain GALNTs were effective at glycosylating S (Fig. 1H). S glycosylation (as detected by HPA reactivity) was detected upon coexpression with GALNT1, 2, 3, 6, 10, 13 or 16 (Fig. 1H). No glycosylation was seen upon coexpression with GALNT4, 5, 7, 8, 9, 11, 12, 14, 15, 17, 18, 19 or 20 (Fig. 1H).

3We next examined which of the identified GALNTs are capable of glycosylating the furin proximal region of S. GALNT1, 2, 3, 6, 13 and 16 were expressed, partially purified and tested for GalNAc transfer (as described previously;) to peptides within the furin proximal region of SARS-CoV-2 S. Enzyme assays demonstrated that only GALNT1 showed detectable transferase activity above background against this region of S (Fig. 2A–B and Fig. S2A–F). Additionally, examining the activity of GALNT1 against peptides where each potential site of O-glycosylation was eliminated revealed a dramatic loss of activity against a peptide containing mutations at T678, suggesting that this is a preferred site of GalNAc addition by GALNT1 (Fig. 2A–B and Fig. S2A). T678 is one of the sites previously identified as being O-glycosylated in human cells by mass spectroscopy. Likewise, a decrease in GalNAc transfer was seen with a peptide containing mutations at T676, suggesting this is also a site of addition for GALNT1 (Fig. 2A–B and Fig. S2A). Interestingly, complete abrogation of all GALNT1-mediated glycosylation was seen upon mutating the unique proline at position 681 (P681), indicating that GALNT1 activity in this region is dependent on this proline. No glycosylation of the comparable region of SARS-CoV-1 (that lacks the novel PRRA insertion) by GALNT1 was observed. Taken together, these results suggest GALNT1 glycosylates specific residues within the furin proximal region of the SARS-CoV-2 S and that glycosylation by GALNT1 is dependent on the presence of P681.

4Our cell culture experiments revealed that O-glycosylation was only present on the full-length S, and never seen on either the S1 or S2 proteolytic fragments (Fig. 1H), suggesting that these events are mutually exclusive. To investigate whether O-glycosylation may interfere with furin cleavage of S, we coexpressed S with various GALNTs and assessed the ratio of cleaved to intact S. S was cotransfected with either GALNT1 (which glycosylates S) or GALNT4 (which does not glycosylate S) and ratios of S were quantitated via western blots. As shown in Fig. 2C–D and Fig. S3, coexpression of S with GALNT1 resulted in O-glycosylation of S and dramatically reduced the degree of furin cleavage. To examine whether O-glycosylation modulates furin cleavage of S in mammalian cells, we performed the same experiments in Vero E6 cells, comparing the effects of GALNT1 to GALNT8 (which does not glycosylate S). As shown in Fig. 2E–G and Fig. S4, coexpression of S with GALNT1 resulted in increased O-glycosylation of intact S and decreased furin cleavage, relative to coexpression with GALNT8. Taken together, our results suggest that O-glycosylation of S by specific members of the GALNT family can modulate furin cleavage and the ratio of cleaved to intact S produced in mammalian cells.

5Given the potential role of endogenously expressed GALNTs to alter the processing of S in cells, we next set out to define the repertoire of GALNTs expressed in cells of the respiratory tract likely to be infected by SARS-CoV-2. We mined the single-cell RNA sequencing databases from cells of the lower respiratory tract of healthy controls from Travaglini et al., 2020. As shown in Fig. 3, GALNT1, 2, 3, 6, 7, 11, 12 and 18 are the predominant family members expressed. In particular, abundant expression of GALNT1 is notable across many cell types that also express ACE2, and are thus likely targets for SARS-CoV-2 infection (Fig. 3A–B). We next examined expression levels of GALNTs in the upper respiratory cells of healthy controls from the dataset of Chua et al., 2020 (Fig. 3C–D). In cells expressing the highest levels of ACE2 (ionocytes), GALNT1 is the most abundantly expressed family member (Fig. 3C–D).

6Our results identify O-glycosylation as a modulator of furin cleavage of the SARS-CoV-2 S protein. Given the role of the furin cleavage in increased infectivity and syncytia formation, our results suggest that O-glycosylation may alter viral tropism and/or infectivity via influencing the degree to which S is processed by furin, raising the possibility that differences in GALNT expression from individual to individual could influence disease severity. In particular, we identify one O-glycosyltransferase, GALNT1, that is expressed within ACE2+ cells of the upper and lower respiratory tracts, and specifically glycosylates the furin proximal region within S. Additionally, the activity of GALNT1 within the furin proximal region is dependent on the unique P681 present in SARS-CoV-2, which is supported by previous studies indicating that GALNT1 has a strong preference for substrates with vicinal prolines. Given that the recently identified B.1.1.7 variant contains mutations in P681, our results suggest that the effect of the loss of that proline on glycosylation of S may be a contributing factor to its increased infectivity.

7The codon optimized cDNA of full-length spike (Genscript) was amplified by PCR and digested by HindIII and NotI, then cloned into pIB/V5-His vector (Invitrogen), fused with V5 epitope at the C-terminus. To make mutations to remove the furin cleavage site (spikeDfurin), two arginines at position 682 and 683 were changed to alanines using the QuickChange II XL site-directed mutagenesis kit (Agilent Technologies) and the plasmid pIB-spike-V5 as a template. Mutagenic primers were designed using the QuickChange Primer Design Program available online. The sense primer sequence is 5’-cgagcacggcgagctgagttagtctggg-3’ and anti-sense primer sequence is 5’-cccagactaactcagctcgccgtgctcg-3’. Reactions and transformation were performed according to the manufacturer’s instructions. DNA sequencing was performed to verify the mutations. The cDNA of hACE2 (Genscript) was amplified by PCR and digested by KpnI and NotI, then cloned into pIB/V5-His vector, fused with mAPPLE/V5 epitope at the C-terminus. The cDNAs of GALNTs 1–20 (Genscript) were amplified by PCR and digested by BamHI and NotI (GALNT1, 2, 3, 4, 9, 13, 14, 17, 18, 19, 20) or HindIII and NotI (GALNT5, 6, 8, 10, 11, 15, 16) or HindIII and XhoI (GALNT12), or SacI and NotI (GALNT7), then cloned into the pIB vector, fused with FLAG epitope at the C-terminus.

8pcDNA3.1-spike and pcDNA-3.1-spikeΔfurin (Genscript) were fused with V5 epitope at the C-terminus to make pcDNA3.1-spike-V5 and pcDNA-3.1-spikeΔfurin-V5.

9The plasmids pIB-spike-V5 with or without pIB-GALNTs-FLAG were transfected to S2R+ cells (DGRC) using Effectene transfection reagent (Qiagen) according to the instructions. For expression of Spike in Vero E6 cells (ATCC), the cells were transfected with pcDNA3.1-spike-V5 with or without pcDNA3.1-GALNTs-FLAG (Genscript) using Lipofectamine 3000 reagent (Invitrogen) according to the instructions.

10For western blotting, cells were lysed with RIPA buffer (Sigma) containing 1X Halt Protease Inhibitor (Thermo Scientific) 3–4 days after transfection. Protein extracts were incubated with agarose-immobolized V5 antibody (Bethyl) overnight at 4°C. Then beads were collected and washed with PBS four times and protein samples were mixed with LDS sample buffer (Invitrogen) containing BME. Proteins were analyzed by NuPAGE 4–12% Bis-Tris gels and transferred onto nitrocellulose membranes. The membranes were blocked with Odyssey Blocking Buffer (PBS-based) (Li-COR), then incubated with anti-V5 antibody (dilution 1:1000, Invitrogen) or anti-S1 antibody (dilution 1:1000, GeneTex) overnight at 4 °C. After washing with PBS containing 0.1% Tween-20 (PBST), the membrane was incubated with IRDye 680LT-conjugated anti-mouse IgG (1:5000, Li-COR) or 680LT-conjugated anti-rabbit IgG (1:5000, Li-COR) and IRDye 800CW (Li-COR) labeled HPA (Helix pomatia lectin, Sigma) or PNA (peanut agglutinin, Sigma) for 1hr at room temperature. After washing with PBST and PBS, the membranes were scanned by a Li-COR Odyssey Infrared Imaging System. The protein band intensity was measured using ImageJ to calculate the ratio of cleaved protein/intact protein and the ratio of HPA/intact protein.

1124 hrs after transfection, the furin inhibitor decanoyl-RVKR-CMK (Tocris) was added to the medium at final concentrations of 20uM, 50uM or 100uM. Cells were collected 72 hrs after transfection and lysed with RIPA buffer (Sigma) containing 1X Halt Protease Inhibitor (Thermo Scientific). The cell lysates were used for immunoprecipitation and western blotting experiments.

12Cells were fixed with 4% PFA, washed with PBST and mounted with Vectashield Mouting Medium with DAPI (Vector Laboratories) 2 days after transfection. The cells were imaged on Nikon A1R confocal microscope. Images were processed using ImageJ.

13Expression of recombinant GALNTs was performed either using Pichia pastoris or COS7 cells as described previously and referenced in the main text. Peptide substrates were synthesized by Peptide 2.0. Assays were performed as described previously and referenced in the main text. Reactions were run for 30 min at 37°C in 25 μl final volumes consisting of the following: 25 mM acceptor substrate, 7.3 μM UDP-[14C]-GalNAc (54.7 mCi/mmol; 0.02mCi/ml), 2 mM cold UDP-GalNAc, 500 mM MnCl2, 40 mM cacodylate (pH 6.5), 40 mM 2-mercaptoethanol and 0.1 % Triton X-100. Reactions were then quenched with 30 mM EDTA. Glycosylated products were separated from unincorporated UDP-[14C]-GalNAc by anion exchange chromatography using AG1-X8 resin columns (Bio-Rad #1401454), and product incorporation was determined by liquid scintillation counting (Beckman Coulter LS6500). Reactions without acceptor peptide were also used to generate background values that were subtracted from each experimental value. Assays for each peptide substrate were run in triplicate and repeated three times. Experimental values for each substrate were then averaged and standard deviations were calculated. Enzyme activity (initial rate) is expressed as CPM/hr/μM or CPM/hr.

14DNase-free RNA was isolated by using PureLink RNA Mini Isolation Kit (Ambion). cDNA synthesis was performed by using the iScript cDNA Synthesis Kit (Bio-Rad). Human primers were designed by using Beacon Designer software (Bio-Rad). Quantitative PCR was performed on a CFX96 real-time PCR thermocycler (Bio-Rad) using the SYBR-Green PCR Master Mix (Bio-Rad). Quantitative PCR was performed in triplicate. Gene expression levels were normalized to human 29S rRNA and plotted as gene expression level and relative expression levels to a housekeeping gene. Values represent mean value.

15Published datasets for upper and lower airways were retrieved as R data format (.rds) and the encapsulated Seurat (v.3.2.3) object was updated to Seurat v3 object using the Seurat UpdateSeuratObject function. We subset the data for the Epithelial cell compartment from the upper and lower airways atlas and the expression of GALNTs and ACE2 is illustrated as dot plots using the R ggplot2 (v.3.3.2) library. Individual dot plots for upper and lower airways were assembled in Adobe Illustrator (v.24.3) to generate a final figure. The datasets are available for download from their respective published sources referenced in the main text.