Identification of β4GALNT2 as an anti-hPIV3 factor through genome-wide CRISPR/Cas9 library screening

Abstract

Human respirovirus 3 (also known as human parainfluenza virus 3; hPIV3) is a major cause of severe acute respiratory infections in vulnerable populations. Here we conducted a genome-wide CRISPR/Cas9 library screen to identify key host factors for hPIV3 infection. In addition to identifying several host proteins involved in glycosylation as proviral factors, we identified β-1,4-N-Acetyl-Galactosaminyltransferase 2 (β4GALNT2) as a potent restriction factor. Further investigation demonstrated that the addition of a GalNAc residue to α2-3-sialylated glycans by β4GALNT2, resulting in the Sd^a glycotope, disrupted the interaction between the viral hemagglutinin-neuraminidase (HN) attachment protein and sialoglycan receptors. Specifically, the additional GalNAc residue interfered with the interaction of residue W371 in HN with sub-terminal glycan moieties. β4GALNT2-mediated Sd^a epitope expression also negatively affected infection by other respiroviruses, with the strongest effect being observed for hPIV3.

Keywords: B4GALNT2; CRISPR/Cas9; Paramyxovirus; hPIV3; sialic acid; virus entry.

Figure 1.

Genome-wide CRISPR/Cas9-based genetic screens in Huh7 cells identifying host factors for hPIV3 infection. (A) Schematic of the CRISPR/Cas9 screen workflow in Huh7 cells. Cas9-expressing Huh7 cells were transduced with lentiviral human sgRNA libraries, then infected with GFP-tagged hPIV3 at multiplicity of infection (MOI) 10 for 20 h. Infected cells were sorted by fluorescence-activated cell sorting (FACS) based on GFP expression to isolate high-GFP or GFP-negative populations. Genomic DNA (gDNA) was extracted and sgRNA sequences were amplified, followed by PCR and sequencing. (B) Gating strategy of hPIV3-infected Huh7-mutant pool, the high-GFP expressing or low-GFP expressing cells were enriched accordingly. Negative control are non-infected cells. (C and D) CRISPR screen results for anti- (C) and pro-(D)-viral genes in Huh7 cells. Genes were ranked by robust rank aggregation (RRA) scores, calculated using MAGeCK [25]. Lower RRA scores in panels C and D indicate antiviral and proviral genes, respectively. (E) STRING network [26] analysis of the top-6 proviral genes (highlighted in red). Genes shown in black were automatically added by selecting the “add more nodes” option to connect the input genes. Colours represent the types of interactions according to the legend. The functions of genes listed in C and D are listed in Table S2.

Figure 2.

Expression of Sd^a antigen by B4GALNT2 inhibits hPIV3 infection. (A) Schematic representation of the glycan modification catalyzed by β4GALNT2. Glycan symbols are drawn according to the SNFG format [34]. (B) Fluorescently labelled DBA or 2–3 Lec were applied to HEK^ΔSia cells co-transfected with ST3GAL4 (0.5 μg) or β4GALNT2 (4 μg)-encoding plasmids, either independently or in combination. Scale bar = 20 μm. (C) Flow cytometry histogram overlays showing the binding of the specific lectins to HEK^ΔSia cells co-transfected with varying amount of ST3GAL4 and β4GALNT2-encoding plasmids. Dolichos biflorus agglutinin (DBA), recognize terminal GalNAc; 2–3 Lectenz (2-3 Lec) recognizes 2–3 sialylated N-glycans. (D) Flow cytometry analysis of different virus infection efficiency (MOI = 2) in HEK^ΔSia cells co-transfected with varying amount of ST3GAL4 and β4GALlNT2-encoding plasmids. Data are normalized to the infection level of HEK^ΔSia cells transfected with only ST3Gal4-encoding (0.5 μg) plasmid. The bars represent the mean of two independent experiments, each performed in triplicate. The flow cytometry analysis gating strategy is presented in Fig. S3. Data are presented as mean ± SD. * P≤0.05, ** P≤0.01, *** P≤0.001 and **** P < 0.0001.

Figure 3.

Sd^a antigen expression reduces paramyxovirus HN and influenza HA binding to glycoprotein receptors. (A) LAMP1 and LAMP1-Sd^a glycoproteins were loaded onto the BLI sensor at comparable levels. (B) Lectin characterization of LAMP1 and LAMP1-Sd^a using DBA and MAL I. MAL I recognizes Neu5Acα2-3Galβ1-4GlcNAc (a trisaccharide common in N-glycans). (C) BLI analysis to determine the binding kinetics of nanoparticles displaying paramyxovirus HN or influenza HA glycoproteins to LAMP1 and LAMP1-Sd^a. Binding was assessed using an established HN-Ni NTA nanoparticle (HN-NPs) system using 3.5×10¹⁰ HN-NPs or HA-NPs per well. Binding of NDV HN-NPs was additionally tested in the presence of site-I specific inhibitor BCX2798 (BCX). Each experiment was conducted at least twice, with representative results displayed here.

Figure 4.

Structural basis for Sd^a-mediated disruption of hPIV3 HN-glycan interactions. (A, B) Predicted interactions between hPIV3 HN monomer and ligands using Chai Discovery (https://lab.chaidiscovery.com) [22]. Panel (A) illustrates the binding of HN to 3′-sialyl-N-acetyllactosamine (3'SLN), while panel (B) depicts the interaction with the Sd^a glycotope. Aromatic residues W371, F372 and W428 are shown in red. (C) Overlay of the predicted interactions, highlighting the differences between the two ligands. The interaction between GalNAc and residues C214 and R129 induces a shift in the entire glycan chain, as indicated by the red arrow. (D) Binding analysis of hPIV3 HN wild-type (WT) and W371A mutant HN-NPs to 3'S(LN)₃ was conducted using BLI, in the presence or absence of the BCX2798 inhibitor, using 3.5×10¹⁰ HN-NPs per well. The difference in the binding curve for hPIV3 HN-NPs compared to Figure 3C results from the different receptors (LAMP1 vs 3’S[LN]₃) being used. All experiments were performed independently in triplicate, with representative data presented.