Genome-Wide CRISPR/Cas9 Screen Reveals a Role for SLC35A1 in the Adsorption of Porcine Deltacoronavirus

ABSTRACT

Porcine deltacoronavirus (PDCoV), an emerging enteropathogenic coronavirus, not only causes diarrhea in piglets but also possesses the potential to infect humans. To better understand host-virus genetic dependencies and find potential therapeutic targets for PDCoV, we used a porcine single-guide RNA (sgRNA) lentivirus library to screen host factors related to PDCoV infection in LLC-PK1 cells. The solute carrier family 35 member A1 (SLC35A1), a key molecule in the sialic acid (SA) synthesis pathway, was identified as a host factor required for PDCoV infection. A knockout of SLC35A1 caused decreases in the amounts of cell surface sialic acid (SA) and viral adsorption; meanwhile, trypsin promoted the use of SA in PDCoV infection. By constructing and assessing a series of recombinant PDCoV strains with the deletion or mutation of possible critical domain or amino acid residues for SA binding in the S1 N-terminal domain, we found that S T182 might be a PDCoV SA-binding site. However, the double knockout of SLC35A1 and amino peptidase N (APN) could not block PDCoV infection completely. Additionally, we found that different swine enteric coronaviruses, including transmissible gastroenteritis coronavirus, porcine epidemic diarrhea virus, and swine acute diarrhea syndrome coronavirus, are differentially dependent on SA. Overall, our study uncovered a collection of host factors that can be exploited as drug targets against PDCoV infection and deepened our understanding of the relationship between PDCoV and SA.

IMPORTANCE Identifying the host factors required for replication will be helpful to uncover the pathogenesis mechanisms and develop antivirals against the emerging coronavirus porcine deltacoronavirus (PDCoV). Herein, we performed a genome-wide clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 knockout screen, the results of which revealed that the solute carrier family 35 member A1 (SLC35A1) is a host factor required for PDCoV infection that acts by regulating cell surface sialic acid (SA). We also identified the T182 site in the N-terminal domain of PDCoV S1 subunit as being associated with the SA-binding site and found that trypsin promotes the use of cell surface SA by PDCoV. Furthermore, different swine enteric coronaviruses use SLC35A1 differently for infection. This is the first study to screen host factors required for PDCoV replication using a genome-wide CRISPR-Cas9 functional knockout, thereby providing clues for developing antiviral drugs against PDCoV infection.

INTRODUCTION

Porcine deltacoronavirus (PDCoV) is a single-stranded, positive-sense, enveloped RNA virus belonging to the genus Deltacoronavirus in the family Coronaviridae of the order Nidovirales (1, 2). PDCoV was first reported in Hong Kong in 2012 (3). However, its pathogenic potential was not recognized until 2014, when it was found to be the cause of piglet diarrheic outbreaks, beginning on a few farms in Ohio, then spreading throughout the United States and becoming a global issue (4). Now, it is one of the most common enteropathogenic viruses in piglets, causing the typical clinical symptoms of vomiting, acute diarrhea, dehydration, and even death, and leading to economic losses in the swine industry (5, 6). Compared with other porcine enteropathogenic coronaviruses, such as transmissible gastroenteritis coronavirus (TGEV), porcine epidemic diarrhea virus (PEDV), and swine acute diarrhea syndrome coronavirus (SADS-CoV), PDCoV possesses the characteristics of wide tissue tropism and cross-species infection (7). Notably, Lednicky et al. (8) recently reported that PDCoV was isolated from plasma samples of three Haitian children with acute undifferentiated febrile illness in 2021, and they speculated that PDCoV had significant cross-species transmission potential and risk of infecting humans, sparking an increased interest in studying this emerging virus.

The specific binding of spike (S) protein to its cell surface receptor is the most critical step of coronavirus (CoV) infection because it mediates virus adsorption and invasion (9, 10). The PDCoV S protein is a type I transmembrane glycoprotein prominent on the surface of virus particles. It is composed of the S1 and S2 subunits; the S1 subunit is responsible mainly for cellular receptor recognition, whereas the S2 subunit is involved in the subsequent membrane fusion (10). Similar to the S1 subunits of other CoVs, PDCoV S1 contains two domains: the N-terminal domain (S1 NTD) and C-terminal domain (S1 CTD). The crystal structure of the PDCoV S1 NTD revealed that it has a sandwich-like structure like that of integrins (a class of proteins that can bind polysaccharides) and can bind sugars on the cell surface, enhancing the cell surface adsorption capacity. The PDCoV S1 CTD has the same structural fold as the α-CoV S1 CTDs, whereas it differs from that of the β-CoV S1-CTDs, and it is a pivotal domain for binding to amino peptidase N (APN) or another unidentified receptor on the host cell surface (11, 12). The cellular receptors of different genera of CoV vary greatly; for example, APN (13, 14), angiotensin-converting enzyme 2 (ACE2) (15–17), dipeptidyl peptidase 4 (DPP4) (18), and carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) (19) have each been identified as a receptor of some α-CoVs and β-CoVs, whereas γ-CoV uses cell surface sialic acid (SA) as a receptor (20). At present, APN has been identified as a putative receptor for PDCoV. Several studies reported that the ectopic expression of APN in nonsusceptible cells, such as BHK-21, HeLa, or Vero cells, could support PDCoV infection. However, the knockout of APN from susceptible cells, such as IPI-2I and ST cells, reduced the PDCoV infection only to a certain extent, suggesting that an APN knockout incompletely blocks PDCoV infection (21, 22). Moreover, two independent research groups reported that APN-knockout pigs are still susceptible to PDCoV infection, suggesting that APN is not the key function receptor for PDCoV infection in vivo (23, 24). Additionally, a recent study reported that APN mediates PDCoV entry via an endocytic pathway (25). All the above-mentioned studies suggested that APN may be one receptor for PDCoV infection but that other unknown receptors or factors assisting or acting independently of APN still exist and play an important role in PDCoV invasion.

Clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 is a powerful technology that has provided an effective method for introducing targeted loss-of-function mutations at specific genome sites in diverse species. Recently, the genome-wide CRISPR knockout (GeCKO) screening strategy has been used to investigate the host factors required for viral infection. For example, multiple independent genome-wide CRISPR single-guide RNA (sgRNA) library screens identified TMEM41B as a host factor required for infection by several CoVs, including severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), human coronavirus (HCoV)-OC43, HCoV-NL63, HCoV-229E, Middle East respiratory syndrome coronavirus (MERS-CoV), and TGEV (26–29). Furthermore, Hoffmann et al. (30) performed a genome-wide loss-of-function CRISPR-Cas9 screen that identified TMEM41B as being required for flavivirus replication. Consistent with the results of these two independent screens, Luo et al. (31) identified zinc finger DHHC-type palmitoyltransferase 17 (ZDHHC17 or ZD17) as an important host factor for SARS-CoV infection. However, there have been no reports of applying GeCKO screening to PDCoV infection.

In this study, we performed two independent GeCKO-based genetic screens in LLC-PK1 cells to identify host factors required for PDCoV infection. The results showed that SLC35A1 is a significant host factor for PDCoV infection. A knockout of SLC35A1 led to a smaller amount of SA on the cell membrane surface as well as markedly less PDCoV attachment. Trypsin promoted use of SA by PDCoV. We also demonstrated that the dependence of different swine enteric coronaviruses (SECoVs) on SA is diverse and that PDCoV can establish valid infection even when both SLC35A1 and APN are knocked out.

RESULTS

Genome-scale CRISPR screen identifies host factors associated with PDCoV infection.

To identify host factors that are necessary for PDCoV infection, we first constructed an LLC-PK1 cell line stably expressing Cas9 (LLC-PK1-Cas9) (Fig. 1A). Two sgRNA lentivirus libraries, “A library” containing 59,088 sgRNAs and “B library” containing 55,007 sgRNAs, were then designed and synthesized to target a total of 20,081 genes in the porcine genome (data not shown). The recombinant lentiviruses expressing these sgRNAs were transduced into LLC-PK1-Cas9 cells, which were subsequently inoculated with PDCoV at a multiplicity of infection (MOI) of 0.1. After five rounds of PDCoV infection, the PDCoV-resistant cells were enriched (Fig. 1B). The genomic DNA (gDNA) of the surviving cells was extracted and subjected to next-generation sequencing (NGS) to determine the enrichment of the integrated sgRNAs. The top 50 identical genes screened from both the A and B libraries were selected for further verification (Fig. 1C). Fifty LLC-PK1 polyclonal knockout cell lines were constructed using the CRISPR/Cas9 gene editing system, each of which was then inoculated with PDCoV to determine their differences in susceptibility according to the cytopathic degree at the same time postinoculation. The results showed that preliminary knockout of transmembrane protein 41B (TMEM41B), solute carrier family 35 member A1 (SLC35A1), sorting nexin 10 (SNX10), VOPP1 WW domain binding protein (VOPP1), or proprotein convertase subtilisin/kexin type 6 (PCSK6) could inhibit the proliferation of PDCoV (Fig. 1D). Because TMEM41B has already been demonstrated to be required for CoV replication (26, 27), and the screening score for SLC35A1 was only slightly lower than that for TMEM41B, SLC35A1 was selected for further analysis.

FIG 1.

Genome-scale CRISPR screening of host factors associated with PDCoV infection. (A) Representative Western blot showing the stable expression of Cas9 protein in LLC-PK1 cells. (B) Schematic diagram illustrating the workflow of sgRNA lentivirus library screening. LLC-PK1 Cas9 cells were transduced with the porcine sgRNA lentivirus libraries A and B, and mutant cells were then infected with PDCoV at an MOI of 0.1. Surviving cells from this virus challenge were isolated, and their genomic DNA (gDNA) was extracted and subjected to sequencing. (C) Enrichment scores of the top 50 host factors in the cross-sorting of libraries A and B. (D) To achieve preliminary screening of the host factors susceptible to PDCoV, LLC-PK1 polyclonal KO cell lines were inoculated with PDCoV. Several molecules that might contribute to PDCoV infection are highlighted; the specific infection of each polyclonal cell line was determined by CPE and IFA. Typical PDCoV-induced CPE (i.e., shrunken cells, syncytia, and exfoliated cells) is indicated by red arrows. PDCoV N protein is presented in green, and blue denotes DNA. Scale bar, 100 μm.

SA facilitates PDCoV infection.

SLC35A1 is one of the key molecules in the SA synthesis pathway on the cell surface, and a previous study suggested that cell surface SA is significantly reduced after the knockout of SLC35A1 (32). To explore the relationship among SA, SLC35A1, and PDCoV, SLC35A1-knockout (SLC35A1-KO) LLC-PK1 and IPI-FX cell lines were constructed, and their successful construction was confirmed via sequencing (data not shown). Consistent with the previous report (32), our SLC35A1 knockout led to significantly less SA on the cell surfaces of LLC-PK1 and IPI-FX cells, as demonstrated by the results of a fluorescein-labeled wheat germ agglutinin (WGA)-binding assay (Fig. 2A). To investigate whether SA is involved in PDCoV infection, wild type (WT) and SLC35A1-KO LLC-PK1 cells were infected with PDCoV (MOI = 0.01). At 12 and 24 h postinfection (hpi), the cytopathic effect (CPE) in SLC35A1-KO cells was muted, whereas evident CPE, characterized by shrunken cells, syncytium, and exfoliated cells, was observed in WT LLC-PK1 cells (Fig. 2B). The results of an indirect immunofluorescence assay (IFA) with antibody against PDCoV nucleocapsid (N) protein showed that the infection rate at 12 hpi was significantly lower in SLC35A1-KO cells than in WT LLC-PK1 cells; however, only relatively small differences between these cell types were observed at 24 hpi (Fig. 2C). The viral titers were consistent with the IFA results (Fig. 2D). Similar results were also observed in both WT and SLC35A1-KO IPI-FX cells (Fig. 2E to G). In addition, we compared the plaque diameters in PDCoV-infected WT and SLC35A1-KO LLC-PK1 cells. The results showed that both the diameter and number of plaques formed by PDCoV were smaller in SLC35A1-KO LLC-PK1 cells than in WT LLC-PK1 cells (Fig. 2H and I). Together, these results suggest that SA is required for PDCoV infection.

FIG 2.

SA facilitates PDCoV infection. (A) SA content on the surface of WT and SLC35A1 KO LLC-PK1 (left) or IPI-FX (right) cells was detected by WGA-binding assay. (B) CPE in mock- or PDCoV-infected WT LLC-PK1 cells and SLC35A1-KO cells at 12 and 24 hpi. Typical PDCoV-induced CPE is indicated by red arrows. (C) PDCoV N protein on WT and SLC35A1-KO LLC-PK1 cells, visualized by IFA. (D) Virus titers of PDCoV-infected WT and SLC35A1-KO LLC-PK1 cells were evaluated by TCID₅₀ assay. (E) CPE in mock- or PDCoV-infected IPI-FX WT cells and SLC35A1-KO cells. Typical PDCoV-induced CPE is indicated by red arrows. (F) PDCoV N protein on WT and SLC35A1-KO IPI-FX cells, visualized by IFA. (G) Virus titers of PDCoV-infected WT and SLC35A1-KO IPI-FX cells were evaluated by TCID₅₀ assay. (H) Plaque assay on PDCoV-infected WT and SLC35A1-KO LLC-PK1 cells. (I) Statistical analysis for the plaques diameter of PDCoV-infected WT and SLC35A1-KO LLC-PK1 cells. In panels C and F, viral N protein is presented in green, and blue denotes DNA. Scale bar, 100 μm. In panels D and G, all data are presented as the mean ± SD of triplicates. Statistical analysis was conducted by an unpaired t test or one-way ANOVA; ns, no significant difference; *, P ≤ 0.05; ****, P ≤ 0.0001.

Trypsin promotes use of cell surface SA by PDCoV.

To investigate whether SA affects PDCoV adsorption, SLC35A-KO and WT LLC-PK1 cells were inoculated with PDCoV (MOI = 10) for 1 h at 4°C, and then the cells were collected for RT-qPCR. As shown in Fig. 3A, the attachment capacity of PDCoV was clearly weakened on the SLC35A1-KO cells compared with that of the WT cells, suggesting that SA affects PDCoV adsorption on cells. Similar results were observed in SLC35A-KO IPI-FX cells (Fig. 3B). These results are consistent with the fact that most glycans, including SA, are used as coreceptors and influence the viral attachment stage.

FIG 3.

Trypsin promotes use of cell surface SA by PDCoV. (A and B) PDCoV attachment on WT and SLC35A1-KO LLC-PK1 (A) or IPI-FX (B) cells. (C and D) PDCoV attachment on WT and SLC35A1-KO LLC-PK1 cells in the presence of 7.5 μg/mL trypsin (C) or on cells maintained with 2% FBS in the absence of trypsin (D). All data are presented as the mean ± SD of triplicates. Statistical analysis was conducted by an unpaired t test; ns, no significant difference; ***, P ≤ 0.001; ****, P ≤ 0.0001.

A previous study showed that PDCoV proliferation requires the involvement of trypsin, and trypsin supports PDCoV replication by promoting cell-to-cell membrane fusion (33). To explore whether a correlation exists among PDCoV, trypsin, and SA, we conducted viral adsorption experiments with or without trypsin. The RT-qPCR results showed that, in the presence of trypsin, the amount of PDCoV attached on WT LLC-PK1 cells was three times higher than that on SLC35A1-KO LLC-PK1 cells (Fig. 3C), whereas in the absence of trypsin, there was no difference between the two groups (Fig. 3D). These results suggest that the use of SA during PDCoV attachment requires the involvement of trypsin.

The S1 NTD α-helix is not a putative SA-binding site for PDCoV.

The crystal structure of the PDCoV S protein was reported previously, and the S1 NTD of CoVs is generally considered to be a glycoside-binding domain (34). In addition, an α-helix structure (composed of 160 to 169 amino acids in the S1 NTD region) at the top of the S1 NTD was recently speculated to be a possible glycoside-binding site for PDCoV (11, 12). To test whether the S1 NTD or the α-helix structure is the SA-binding site, we initially attempted to construct recombinant PDCoV mutants in which the S1 NTD or the α-helix structure was deleted using the established reverse genetics system (RGS) of PDCoV strain CHN-HN-2014 (Fig. 4A). Although we tried various methods, unfortunately, such recombinant viruses could not be successfully rescued, indicating that the entire S1 NTD or α-helix structure may be essential for PDCoV replication.

FIG 4.

The S1 NTD α-helix and H148 are not putative SA-binding sites for PDCoV. (A) Schematic diagram of the construction schemes for the various rPDCoV mutant strains used in our study. Black or red text, respectively, indicate the unsuccessful or successful recovery of rPDCoV mutants via a reverse genetics system. (B) Molecular dynamics simulation of rPDCoV S1 NTD and rPDCoV mut-α; the α-helix is shown with a red border. (C) Identification of different rPDCoV mutants via IFA on LLC-PK1 cells. (D) IFA for the infection of rPDCoV-WT and various PDCoV mutants on WT and SLC35A1-KO LLC-PK1 cells, using anti-PDCoV N protein antibodies. In panels C and D, viral N protein is presented in green, and blue denotes DNA. Scale bar, 100 μm.

Through a molecular dynamics simulation, we found that mutations of the amino acid residues Y162, T164, and N168 can destroy the α-helix structure (Fig. 4B). Based on that finding, we attempted to construct a recombinant PDCoV with Y162P/T164P/N168P mutations. This recombinant PDCoV, termed rPDCoV-mut-α, was successfully rescued, as demonstrated by evident IFA (Fig. 4C), CPE, and a sequence analysis of the S gene (see Fig. S1A to C in the supplemental material). To explore whether the α-helix is the key region for PDCoV binding to SA, WT and SLC35A1-KO LLC-PK1 cells were infected with WT rPDCoV (rPDCoV-WT) or rPDCoV-mut-α (MOI = 0.1), and the virus infection rates in the two cell lines were taken to reflect the use of SA by the virus. The results showed that, like rPDCoV-WT, rPDCoV-mut-α had differing infection efficiencies in the WT and SLC35A1-KO cells, and the use of SA was not abolished in either cell type (Fig. 4D), suggesting that the α-helix structure is not the region where PDCoV binds SA.

S1 NTD T182, but not H148, is a putative SA-binding site for PDCoV.

We further analyzed the protein structure of the adaptive evolution sites and found three positive-selection codons (L106, H148, and T182) in the S1 NTD of PDCoV strain CHN-HN-2014, which was consistent with previous studies (34, 35). To determine whether S protein H148 is a possible SA-binding site for PDCoV, three rPDCoVs with the mutations H148A, H148D, or H148K, respectively, were successfully constructed (Fig. 4C, Fig.S1A to C). WT and SLC35A1-KO LLC-PK1 cells were each infected with rPDCoV-WT or one of the three rPDCoV-mutants (H148A, H148D, or H148K) (MOI = 0.1). The results showed that all three mutants exhibited infection rates like that of the rPDCoV-WT in WT or SLC35A1-KO LLC-PK1 cells (Fig. 4D), indicating that the H148 site is not the putative SA-binding site for PDCoV.

Threonine (T) 182 is located in a loop between two β-sheets in the PDCoV S1 NTD, and its main chain atoms form hydrogen bonds with T185, which might stabilize the loop, suggesting that T182 is also a candidate site for SA binding (35). Thus, we constructed a rPDCoV in which the threonine (T) at 182 was substituted with alanine (A) (rPDCoV-S-T182A). Evident CPE could be observed in LLC-PK1 cells infected with the recused rPDCoV-S-T182A (Fig. S1D), and the sequence analysis (Fig. S1E), results of a PCR test (Fig. S1F), and IFA (Fig. 5A) also confirmed that rPDCoV-S-T182A was successfully rescued. We then infected WT and SLC35A1-KO LLC-PK1 cells with rPDCoV-WT or rPDCoV-S-T182A. As shown in Fig. 5B, rPDCoV-S-T182A had similar infection rates in WT and SLC35A1-KO LLC-PK1 cells, whereas rPDCoV-WT had a markedly lower infection rate in SLC35A1-KO LLC-PK1 cells than in WT LLC-PK1 cells, indicating that the S T182A mutation could reduce the SA-binding ability of rPDCoV. Thus, S T182 might be a putative SA-binding site. To further test this possibility, we performed adsorption experiments for the rPDCoV-WT and rPDCoV-S-T182A strains on WT and SLC35A1-KO LLC-PK1 cells. The results demonstrated that the adsorption capacity of rPDCoV-S-T182A to SA was attenuated, but not abolished, compared with that of rPDCoV-WT (Fig. 5C and D).

FIG 5.

S1 NTD T182 is putative SA-binding site for PDCoV. (A) Identification of rPDCoV-S-T182A via IFA on LLC-PK1 cells. (B) IFA for the infection of rPDCoV-WT and rPDCoV-S-T182A mutant on WT and SLC35A1-KO LLC-PK1 cells, using anti-PDCoV N protein antibodies. Scale bar, 200 μm. (C and D) rPDCoV-WT (C) or rPDCoV-S-T182A (D) attachment on WT and SLC35A1-KO LLC-PK1 cells. (E) Growth curves of rPDCoV-WT and rPDCoV-S-T182A on LLC-PK1 cells. (F) Plaque assay of rPDCoV-WT and rPDCoV-S-T182A on LLC-PK1 cells. (G) Statistical analysis for the plaques diameter of rPDCoV-WT and rPDCoV-S-T182A on LLC-PK1 cells. In panels C and D, all data are presented as the mean ± SD of triplicates. ns, no significant difference; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.

We next attempted to visualize differences in the proliferation dynamics between the mutant rPDCoV-S-T182A and rPDCoV-WT. To this end, LLC-PK1 cells were infected with rPDCoV-WT or rPDCoV-S-T182A at the same infection dose (MOI = 0.1), and the resulting viral titers were determined at 6, 12, 18, 24, and 30 hpi by performing a 50% tissue culture infective dose (TCID₅₀) assay. As shown in Fig. 5E, the titers of rPDCoV-S-T182A steadily increased within 30 hpi, whereas the titers of rPDCoV-WT reached a peak at 18 to 24 hpi and gradually decreased from 24 to 30 hpi, indicating that the proliferation of rPDCoV-S-T182A has a hysteresis compared with that of rPDCoV-WT. Additionally, the results of viral plaque assays show that the mean diameter of the plaques formed by rPDCoV-S-T182A was much smaller than that of the plaques formed by rPDCoV-WT (Fig. 5F and G), indicating that the replication capacity of rPDCoV-S-T182A was inferior to that of rPDCoV-WT. Collectively, these results demonstrate that S1 NTD T182 is a putative SA-binding site for PDCoV.

The double knockout of SLC35A1 and APN does not completely block PDCoV infection.

Previous studies have demonstrated that APN can act as a receptor to mediate PDCoV entry (21–23); however, PDCoV is still capable of infecting APN-knockout cell lines and APN-knockout gene-edited pigs (23, 24), indicating that other receptors or coreceptors for mediating PDCoV infection must exist. Because SA plays an important role in the attachment stage of PDCoV infection, we tried to further evaluate the significance of SA and APN in PDCoV infection. From the SLC35A1-KO LLC-PK1 cell line, we further knocked out APN, and our sequencing results (Fig. 6A) indicate that we successfully generated a cell line with the double knockout of SLC35A1 and APN, termed SLC35A1/APN-DKO. Previous studies have demonstrated that APN is the key function receptor for TGEV infection, and APN knockout abolished TGEV infection completely (14). As expected, TGEV could not infect the SLC35A1/APN-DKO cells, whereas efficient infection by TGEV could be achieved in the SLC35A1-KO cells (Fig. 6B). WT, SLC35A1-KO, APN-KO, and SLC35A1/APN-DKO LLC-PK1 cells were then each infected with TGEV or PDCoV (MOI = 0.1) and subjected to IFA and TCID₅₀ detection. The results show that TGEV did not infect the APN-KO or SLC35A1/APN-DKO cells, and the infection rate in the SLC35A1-KO LLC-PK1 cells was lower than that in the WT LLC-PK1 cells. However, the PDCoV infection rates in APN-KO and SLC35A1-KO cells were both slightly inferior to that in WT LLC-PK1 cells, and the lowest infection rate was observed in SLC35A1/APN-DKO cells (Fig. 6C). The TCID₅₀ assay results further confirm the PDCoV infection efficiency differences among these four cell lines (Fig. 6D). It should be noted that PDCoV could still infect the SLC35A1/APN-DKO cell line, indicating that, in addition to APN and SA, at least one other unknown receptor for mediating PDCoV infection must exist.

FIG 6.

Double knockout of SLC35A1 and APN does not completely block PDCoV infection. (A) Gene targeting of APN in SLC35A1-KO LLC-PK1 cells using CRISPR-Cas9. Sequencing validation of the SLC35A1/APN-DKO cell line. (B) IFA for the detection of TGEV infection of LLC-PK1 WT, SLC35A1-KO, and SLC35A1/APN-DKO cells using anti-TGEV N protein antibodies. (C) IFA for the detection of PDCoV and TGEV infection on WT, SLC35A1-KO, APN-KO, and SLC35A1/APN-DKO LLC-PK1 cell lines using anti-PDCoV N protein and anti-TGEV N protein antibodies. (D) The virus titers of PDCoV in the above-mentioned cell lines were measured by TCID₅₀ assay. The specific genes lacking in the mutants are listed along the x axis. Viral N protein is presented in green, and blue denotes DNA. Scale bar, 100 μm. In panel D, all data are presented as the mean ± SD of triplicates. *, P ≤ 0.05; ***, P ≤ 0.001.

Different SECoVs have different sensitivities to SA.

We further examined the use of SA by several other SECoVs, including PEDV, TGEV, and SADS-CoV. TGEV-ΔS1-NTD was used as a positive control. We infected WT and SLC35A1-KO LLC-PK1/IPI-FX cells with these viruses, and the infection rates were determined by IFA and TCID₅₀ assay. The results showed that the use of SA by TGEV and SADS-CoV were like that of PDCoV, and the infection capacities of TGEV and SADS-CoV on SLC35A1-KO cells were lower than that on WT cells (Fig. 7A and C). SA affected the TGEV and SADS-CoV infections mainly in the early stage and was usually dispensable (Fig. 7A and C). The infection rates of TGEV-ΔS1-NTD on WT and SLC35A1-KO LLC-PK1 cells were similar (Fig. 7D). Interestingly, we found that the use of SA by PEDV was strain-dependent; the infection of PEDV strain AJ1102, an emerging variant since 2010, was highly SA dependent, whereas the infection of PEDV strain JS2008, a classical PEDV strain, was independent of SA (Fig. 7B and E).

FIG 7.

Different SECoVs have different levels of sensitivity to SA. (A to C) The infection of TGEV (A), PEDV-AJ1102 (B), or SADS-CoV (C) in WT and SLC35A1-KO LLC-PK1 (left) or IPI-FX (right) cells was detected by IFA and TCID₅₀ assay. (D) The infection of TGEV and TGEV-ΔS1-NTD in WT and SLC35A1-KO LLC-PK1 cells was detected by IFA. (E) The infection of PEDV-JS2008 in WT and SLC35A1-KO LLC-PK1 cells was detected by IFA. Viral N or S protein is presented in green, and blue denotes DNA. Scale bar, 100 μm. In panels A, B and C, all data are presented as the mean ± SD of triplicates. Statistical analysis was conducted by one-way ANOVA; ns, no significant difference; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.01; ****, P ≤ 0.0001.

DISCUSSION

PDCoV is an emerging enteropathogenic swine virus, and its potential for cross-species transmission makes understanding its infection mechanism particularly important for human health. Unfortunately, the previous research on PDCoV has not been in-depth and systematic, especially regarding the aspects of receptor discovery and investigations of the infection and replication mechanisms. In the present study, by performing a genome-wide CRISPR-Cas9 knockout screen, we revealed SLC35A1 as a host factor required for PDCoV infection that acts by regulating cell-surface SA. We also demonstrated that SA facilitates PDCoV infection by enhancing PDCoV adsorption. During this process, trypsin promotes the use of SA by PDCoV. Additionally, we determined that the PDCoV S1 NTD α-helix and H148 are not putative SA-binding sites for PDCoV, but S1 NTD T182 is a putative SA-binding site for PDCoV. However, the S T182A mutation did not completely abolish the use of SA by PDCoV. Furthermore, PDCoV was still able to infect SLC35A1/APN-DKO cell lines, suggesting that there are still unknown receptors or c-receptors for mediating PDCoV infection.

Genome-wide CRISPR-Cas9 gene disruption screening is a powerful method for discovering the host factors required for viral infection. Our screening results also identified TMEM41B, which has been reported as a host factor that contributes to the formation of CoV and flavivirus replication organelles, such as double-membrane vesicles (DMVs) (26, 30). As a scramblase (36, 37), TMEM41B is essential for autophagosome formation (38–41), lipid droplet homeostasis (38, 42), membrane contact (43), embryonic development (44), and lipoprotein secretion (42). Recently, multiple independent genome-wide CRISPR sgRNA library screens identified TMEM41B as a pan-CoV host factor, involved in the infections of CoVs, including SARS-CoV-2 (27, 45), HCoV-OC43, HCoV-NL63, HCoV-229E (28), MERS-CoV (46), and TGEV (26), as well as a pan-flavivirus host factor, involved in the infections of flaviviruses, including West Nile virus, yellow fever virus, dengue virus, and Zika virus (30). Given that TMEM41B acts as a pan-CoV host factor, we hypothesized that TMEM41B likely employs the same mechanism for promoting PDCoV infection; thus, we did not study this molecule further. SLC35A1 has been confirmed to play crucial roles in the synthesis of SA that favors PDCoV infection (47, 48). Thus, SLC35A1 was selected for in-depth research. In addition to SLC35A1 and TMEM41B, SNX10, VOPP1, and PCSK6 have also been identified as host factors required for PDCoV infection. SNX10 is a member of the sorting nexin family of proteins that plays crucial roles in cargo sorting in the endosomal pathway via its binding to phosphatidylinositol-3-phosphate (PI3P) localized in early endosomes (49). VOPP1 has been previously shown to colocalize or partially colocalize with cellular vesicular structures, such as endocytosis and autophagy vesicles and perinuclear lysosomes, suggesting that VOPP1-containing vesicles enter final common pathways of the lysosomal system (50). Therefore, we speculate that SNX10 and VOPP1 might act on the internalization or uncoating stage of PDCoV infection. PCSK6 is a PCSK protease family member with unknown functions. Recently, it was reported that PCSK6 is a key protease in the control of smooth muscle cell function in vascular remodeling (51). However, the relationship between PCSK6 and PDCoV has not been reported and deserves further study. Previous studies have identified APN as a receptor for PDCoV infection (21, 22). In our genome-wide CRISPR knockout library screening, APN gene was also hit, but it was ranked backward and not in the top 50 ranked genes in both library A and library B. Based on the sgRNA enrichment score, APN was ranked 191th (score = 1.42), while SLC35A1 was ranked 2nd (score = 1.465). Furthermore, we found that the replication levels of PDCoV were lower in SLC35A1-KO cells than in APN-KO cells (Fig. 6C), consistent with the screening scores of APN and SLC35A1.

SLC35A1 is a key molecule in the SA synthesis pathway. CMP-SA synthetase (CMAS) catalyzes the formation of CMP-SA in the nucleus (52). Subsequently, CMP-SA is transported to the cytoplasm, where it is recognized by SLC35A1 and transported to the Golgi for a series of processing and modification steps to produce glycoconjugates (53). These glycoconjugates are shipped to the cell membrane (54). Therefore, SLC35A1 deficiency will cause lower rates of SA synthesis and transport of SA, leading to a lower level of membrane surface SA. A recent study demonstrated that the anti-S1 antibody with neutralization activity could block PDCoV infection partially through SA binding activity (47). Another study showed that the removal of SA on the cellular surface significantly decreased PDCoV infection, demonstrating that SA is an attachment receptor for PDCoV (48). In this study, we demonstrated that PDCoV can recognize cellular SA and that it uses SA as a coreceptor. It was previously reported that trypsin promotes PDCoV replication by enhancing cell-to-cell membrane fusion (33). Thus, PDCoV requires the addition of exogenous trypsin to mimic the intestinal environment during its isolation and passage. Compared with other tissues, the intestine is rich in trypsin and SA, which may be the main reason why PDCoV infects predominately the small intestine. Here, we further demonstrated that trypsin treatment enhances the ability of PDCoV to bind SA. Interestingly, PDCoV still differentially infected WT and SLC35A1-KO LLC-PK1 cells without trypsin (Fig. S2A to C), suggesting that PDCoV infection still uses SA in the absence of trypsin. The reason for this phenomenon may be that trypsin promotes the use of SA by PDCoV or that SA influences PDCoV infection through mechanisms other than adsorption. A previous study described a two-step process in which MERS-CoV first adheres to SA and then requires subsequent engagement with protein receptors during infectious cell entry but found that SA sufficiently facilitated the later stages of virus spread through cell-cell membrane fusion, without requiring protein receptors (55). Whether SA promotes PDCoV infection through a similar mechanism needs to be further studied. To explore whether the formation of viral syncytium requires the involvement of trypsin or SA, we conducted a further study using TGEV as a model. We infected WT and SLC35A1-KO LLC-PK1 cells with TGEV or TGEV-ΔS1-NTD strains under conditions with 2% fetal bovine serum (FBS) or 7.5 μg/mL trypsin. The results showed that the TGEV and TGEV-ΔS1-NTD strains each formed syncytia in both WT and SLC35A1-KO LLC-PK1 cells in the presence of 7.5 μg/mL trypsin, but they did not form syncytia under the condition with 2% FBS (Fig. S2D). We interpret this to mean that the trypsin environment, but not SA, affected syncytium formation.

Although SA and APN are both coreceptors for PDCoV, their functions in PDCoV proliferation might be different. In this study, we compared the PDCoV infection rates and virus titers on WT, SLC-35A1-KO, APN-KO, and SLC35A1/APN-DKO LLC-PK1 cells. The results showed that SA was more important than APN for PDCoV infection because the SA knockout cells had a lower viral titer. This difference may be a consequence of APN and SA acting on different stages of PDCoV infection. SA acts as an attachment receptor for PDCoV invasion and facilitates its infection, whereas APN mediates PDCoV entry via an endocytic pathway. Furthermore, we examined the use of SA by different SECoVs. The results demonstrated that the use of SA by TGEV and SADS-CoV were like that of PDCoV and that the role of SA use in infection by these viruses is dispensable. Distinctively, we found that PEDV uses SA in a strain-dependent manner; the variant strain AJ1102 was highly SA dependent, whereas the classical strain JS2008 did not use SA. This may be because of the diversity in trypsin dependence between classical and variant PEDV strains. The differences between the classical and variant PEDV strains in their use of SA deserve deeper study.

Recently, PDCoV was shown to exhibit broad host-cell range and tissue tropism. We speculate that this may be due to the multireceptor-mediated viral infection characteristics of PDCoV. Currently, APN orthologues from pigs (pAPN) has been identified as an entry receptor for PDCoV. APN orthologues from other species, including humans (hAPN), mice (mAPN), chickens (chAPN), and felines (fAPN), can also allow PDCoV cell entry (47). Furthermore, TGEV infection was completely blocked in APN-knockout cells and APN/CD163-DKO pigs, whereas PDCoV infection was only slightly decreased (24). In addition, PDCoV can infect SLC35A1/APN-DKO cells, suggesting that, in addition to APN and SA, there are unknown receptors that also contribute to PDCoV infection.

The cryo-electron microscopy structure of the PDCoV S protein with a resolution of 3.3-Å or 3.5-Å has been reported (11, 12). The trimeric protein contains three receptor-binding S1 subunits that tightly pack into a crown-like structure and three membrane fusion S2 subunits that form a stalk (11, 12). The sugar-binding site in the PDCoV S1 NTD is currently unknown but is predicted to be located on top of its core structure (α-helix). H148, which has the highest variability between lineages, is present at the top of the S protein. T182 is located in a loop between two β-sheets of the S protein. Its main chain atoms form hydrogen bonds with S T185, which might stabilize the loop. This study investigated possible SA sites that had been suggested previously. We found that neither the α-helix (160 to 169aa) at the top of the S1 NTD nor S H148 are SA-binding sites of PDCoV, whereas S T182 is a potential SA-binding site of PDCoV. However, although the S T182A mutation lessens the use of SA by PDCoV, it does not completely abolish it. Therefore, there must be other SA-binding sites for PDCoV. It was reported that the SARS-CoV-2 S protein interacts with both cellular heparan sulfate and ACE2 through its receptor-binding domain (56). Docking studies suggest the presence of a heparin/heparan sulfate-binding site adjacent to the ACE2-binding site (56). Accordingly, future work could search for possible SA-binding sites of PDCoV by using similar methods.

In conclusion, we performed a genome-wide CRISPR knockout library screening on PDCoV in LLC-PK1 cells, which identified SLC35A1 as a significant host factor required for PDCoV infection that acts by regulating cell-surface SA, the sugar receptor for PDCoV. Trypsin promotes the use of SA by PDCoV. However, further investigation is needed to determine the main receptors (besides APN and SA) that mediate PDCoV infection. Notably, we verified S T182 as a possible SA-binding site of PDCoV S1, even though mutation to this site does not completely abolish the use of SA by PDCoV.

MATERIALS AND METHODS

Cells and viruses.

LLC-PK1, IPI-FX, and Vero cells were cultured in MEM and DMEM (HyClone and Gibco) supplemented with 10% FBS (NEWZERUM) and 1% penicillin-streptomycin solution (HyClone) at 37°C with 5% CO₂. PDCoV strain CHN-HN-2014 (GenBank accession number KT336560), SADS-CoV strain CHN-GD-2017 (GenBank accession number MH539766), and PEDV strain AJ1102 (GenBank accession number JX188454.1) were propagated and titrated on LLC-PK1 and Vero cells in MEM and DMEM supplemented with 7.5 μg/mL trypsin. PEDV strain JS2008 (GenBank accession number KC109141) and TGEV strain WH1 (GenBank accession number HQ462571) were propagated and titrated on ST and Vero cells in DMEM supplemented with 2% FBS. LLC-PK1, IPI-FX, and gene-edited cells were inoculated with the indicated SECoVs and then maintained in the presence of 7.5 or 2.5 μg/mL trypsin or 2% FBS.

LLC-PK1-GeCKO library generation and PDCoV screen.

The LLC-PK1-GeCKO library was generated using the lentiGuide-Puro two-vector system for Cas9 and sgRNA delivery. First, an LLC-PK1 cell clone stably expressing the Cas9 components (LLC-PK1-Cas9) was generated via lentivirus transduction of the Cas9 transgene. We then transduced these cells with a lentivirus porcine gRNA library targeting 20,081 genes, using an MOI of 0.2 to ensure that most cells received only one viral construct. After 48 h, the transduced LLC-PK1-Cas9 cells were selected via culture with 2 μg/mL puromycin for 2 to 3 days. These cells were then infected with PDCoV at an MOI of 0.1 and washed with serum-free medium at 24 to 30 hpi to remove dead cells, before fresh medium was added to the surviving clones. These cells were cultured in MEM supplemented with 10% FBS and 1% penicillin-streptomycin solution and used for subsequent PDCoV infection rounds. The cells were resistant to this viral infection after five rounds of infections. At the end of the screen, the surviving resistant cells were expanded and subjected to deep sequencing analysis.

Generation of knockout cell lines.

A PX459 vector plasmid containing sgRNA-1 (5′-TGGTAATCTGGGTAGATTCA-3′) and sgRNA-2 (5′-GTGGAGTCATACTTGTACAG-3′) targeting the SLC35A1 gene was constructed and used to transduce LLC-PK1 and IPI-FX cells, and another vector PX459 plasmid containing sgRNA-3 (5′-GTAGGCGGTACCGGTTCCA-3′) and sgRNA-4 (5′-CGTTGTGGGTAGGCGGTAC-3′) targeting the APN gene was constructed and used to transduce SLC35A1-KO LLC-PK1 cells; transduction was followed by selection via culture with 3 μg/mL puromycin for 2 days. Surviving cells were cultured in 10% FBS medium. To obtain homogeneous cell lines, LLC-PK1, IPI-FX, and LLC-PK1 SLC35A1-KO cells were subcloned by performing finite continuous dilution in 96-well plates for clonal expansion. Finally, the successful construction of SLC35A1-KO and SLC35A1/APN-DKO cell lines was validated by DNA sequencing.

Flow cytometry.

WT and SLC-35A1-KO cells were detached from tissue-culture plates, centrifuged at 300 × g to form a pellet, washed twice with phosphate-buffered saline (PBS), and maintained at 4°C. Cells were adsorbed with 0.005 mg/mL fluorescein-labeled wheat germ agglutinin (WGA; Sigma-Aldrich, Saint Louis, MO, USA) at 4°C for 60 min, and unbound lectin was removed by washing the cells twice with PBS. The fluorescence intensity of the cell surface was analyzed by flow cytometry.

Viral RNA extraction and quantitative RT-PCR.

Total viral RNA was extracted from PDCoV-infected cells at different time points by using TRIzol reagent (Invitrogen, Madison, WI, USA). Specific primers for the PDCoV N gene (forward, 5′-AGCTGCTACCTCTCCGATTC-3′; reverse, 5′-ACATTGGCACCAGTACGAGA-3′) were designed in accordance with a reference sequence (GenBank accession number KT336560). A real-time one-step quantitative RT-PCR assay performed with the HiScript II one-step qRT-PCR SYBR green kit (Vazyme, China) was used to determine the PDCoV genomic RNA.

Indirect immunofluorescence assay.

LLC-PK1 and IPI-FX cells seeded in 24-well plates were inoculated with the indicated SECoV when the cells reached 100% confluence. At 12 and 24 hpi, CPE could be observed by optical microscope. The cells were then fixed with 4% paraformaldehyde for 15 min at room temperature and then permeated with methyl alcohol for 10 min. After being washed with PBS, the cells were sealed with PBS containing 5% bovine serum albumin for 1 h, followed by incubation separately with an antibody against PDCoV N, TGEV N, PEDV N, or SADS-CoV S for 1 h at 37°C. The cells were subsequently treated with secondary antibody Alexa Fluor 488-conjugated goat antimouse IgG for 1 h at 37°C, followed by treatment with 4′, 6-diamidino-2-phenylindole (DAPI) (Beyotime, Shanghai, China) for 10 min at room temperature. After the samples were washed with PBS, fluorescent images were examined with a fluorescence microscope (Olympus).

TCID₅₀ and plaque assays.

The titers of PDCoV and TGEV in WT and KO cells were determined by TCID₅₀ and plaque assay on LLC-PK1 cells; the titers of PEDV and SADS-CoV were determined by TCID₅₀ assay on Vero cells. The infected cell samples were collected and serially diluted 10-fold; 100 μL of each dilution was added into a single well of a 96-well plate, and 800 μL of each dilution was added into a single well of a six-well plate. After 2 to 3 days of incubation at 37°C with 5% CO₂, viral CPE was monitored, and the viral titers were calculated by using the Reed-Muench method and expressed as TCID₅₀ per milliliter. Plaques were observed and then stained with 0.1% crystal violet.

Western blot.

Cells were washed three times with PBS and then harvested on ice with radioimmunoprecipitation assay (RIPA) lysis buffer (Beyotime). The resulting lysates were centrifuged and then boiled at 100°C in 5× SDS-PAGE loading buffer for 10 min. Next, the protein samples were separated by SDS-PAGE and electroblotted onto polyvinylidene difluoride membranes (Millipore, Burlington, MA, USA). These membranes were blocked with Tris-buffered saline with Tween 20 (TBST) supplemented with 10% skim milk and incubated with the indicated antibodies at room temperature, followed by a second incubation with horseradish peroxidase (HRP)-conjugated antimouse IgG antibodies or HRP-conjugated antirabbit IgG antibodies. Finally, the protein was detected via the use of a chemiluminescent substrate (Bio-Rad, USA).

Cleavage of the pBAC-rPDCoV and construction of the recombinant BAC.

The specific cleavage reaction was conducted in a 50-μL mixture containing 5 μg of pBAC-CHN-HN-2014, 5 μL of Cas9 nuclease (NEB), 20 μg sgRNA (10 μg for each sgRNA), and 5 μL of 10× NEB buffer 3.1 at 37°C for 2.5 h. The cleaved pBAC-CHN-HN-2014 was purified with a DNA Cycle Pure kit (Omega Bio-tek) and verified by electrophoresis in a 0.8% agarose gel. The recombinant pBAC-rPDCoV was constructed via homologous recombination using an In-Fusion cloning kit (Clontech) in a mixture containing the cleaved pBAC-rPDCoV and a modified S gene.

Statistical analyses.

All experiments were performed at least three independent times, and statistical differences were determined by performing Student's t test or one-way ANOVA using GraphPad Prism 6.0 software. In the presented figures, significant differences are indicated by asterisks; ns, no significant difference; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.