Enterovirus D68 receptor requirements unveiled by haploid genetics

Significance

Enterovirus D68 (EV-D68) is an emerging pathogen that recently caused a large outbreak of severe respiratory disease in the United States and is associated with cases of paralysis. Little is known about EV-D68 host factor requirements. Here, using a genome-wide knockout approach, we identified several genes in sialic acid (Sia) biology as being essential for infection. We also showed that not only α2,6-linked Sia, which mainly occurs in the upper respiratory tract, but also α2,3-linked Sia, which mainly occurs in the lower respiratory tract, can serve as the receptor. Moreover, we identified recent EV-D68 isolates that can use an alternative, nonsialylated receptor. Our findings are essential to understand tropism and pathogenesis of EV-D68 as well as the potential of using Sia-targeting inhibitors to treat EV-D68 infections.

Abstract

Enterovirus D68 (EV-D68) is an emerging pathogen that can cause severe respiratory disease and is associated with cases of paralysis, especially among children. Heretofore, information on host factor requirements for EV-D68 infection is scarce. Haploid genetic screening is a powerful tool to reveal factors involved in the entry of pathogens. We performed a genome-wide haploid screen with the EV-D68 prototype Fermon strain to obtain a comprehensive overview of cellular factors supporting EV-D68 infection. We identified and confirmed several genes involved in sialic acid (Sia) biosynthesis, transport, and conjugation to be essential for infection. Moreover, by using knockout cell lines and gene reconstitution, we showed that both α2,6- and α2,3-linked Sia can be used as functional cellular EV-D68 receptors. Importantly, the screen did not reveal a specific protein receptor, suggesting that EV-D68 can use multiple redundant sialylated receptors. Upon testing recent clinical strains, we identified strains that showed a similar Sia dependency, whereas others could infect cells lacking surface Sia, indicating they can use an alternative, nonsialylated receptor. Nevertheless, these Sia-independent strains were still able to bind Sia on human erythrocytes, raising the possibility that these viruses can use multiple receptors. Sequence comparison of Sia-dependent and Sia-independent EV-D68 strains showed that many changes occurred near the canyon that might allow alternative receptor binding. Collectively, our findings provide insights into the identity of the EV-D68 receptor and suggest the possible existence of Sia-independent viruses, which are essential for understanding tropism and disease.

The genus Enterovirus of the family Picornaviridae contains many important pathogens for humans and animals. This genus consists of 12 species: four human enterovirus species (EV-A, -B, -C, and -D), five animal enterovirus species, and three human rhinovirus species. The best known human enterovirus is poliovirus (EV-C), the cause of poliomyelitis and acute flaccid paralysis. Other well-known enteroviruses are the coxsackieviruses (EV-B and EV-C)—which are the main cause of viral meningitis, conjunctivitis, myocarditis, and herpangina—and enterovirus A71, which causes hand-foot-and-mouth disease and is also associated with severe neurological disease, causing serious public health concerns in Southeast Asia (1).

Another emerging enterovirus that causes growing public health problems is enterovirus D68 (EV-D68, a member of the species EV-D). Unlike most enteroviruses, which are acid-resistant and multiply in the human gastrointestinal tract, EV-D68 is an acid-sensitive enterovirus (2) that replicates in the respiratory tract. EV-D68 was first isolated from children with respiratory infections in California in 1962 (3). It was long considered a rare pathogen, but the frequency of detecting EV-D68 during outbreaks of respiratory disease has increased (4, 5) and over the past decades, three clades of EV-D68 (A, B, and C) have emerged and spread worldwide (6, 7). EV-D68 infections mostly cause mild respiratory disease but can also result in severe bronchiolitis or pneumonia, especially among children (4, 5). In 2014, a nationwide EV-D68 outbreak in the United States was associated with severe respiratory disease and a cluster of acute flaccid myelitis and cranial nerve dysfunction in children, implicating EV-D68 as an emerging public health threat (8, 9).

Enteroviruses are small, nonenveloped viruses that contain a single-stranded RNA genome of positive polarity. To initiate infection, enteroviruses bind to specific receptors on host cells. To date, most known enterovirus receptors are cell surface proteins, many of which belong to the Ig superfamily or the integrin receptor family (10). A majority of these receptors bind to the “canyon,” a depression on the virion surface, thereby destabilizing virions and initiating uncoating (11). In EV-D68, the canyon is unusually shallow and narrow, possibly excluding use of large protein receptors (12). Both sensitivity of EV-D68 infection to neuraminidase (NA) treatment and hemagglutination assays point to the use of Sia as the receptor (13, 14). However, beside the role of a terminal Sia residue on the receptor, little is known about the type(s) of Sia that can be used by EV-D68 to infect cells, the composition of the underlying glycan, and whether specific sialylated proteins or glycolipids are required for infection.

Genome-wide genetic screening in human haploid cells is a powerful tool to reveal host factors involved in entry of various pathogens, including viruses (15, 16). In this study, we performed a haploid screen and demonstrate that genes involved in synthesis of sialylated glycans are essential for EV-D68 infection. Furthermore, we show that EV-D68 is able to use α2,6-linked as well as α2,3-linked Sia as a cellular receptor, and we provide the first insights into the composition of the underlying sugar backbone. Finally, we report the identification of recent EV-D68 isolates that can infect Sia-deficient cells, indicating that these strains can use an alternative receptor.

Results

Multiple Genes Involved in Sia Biology Determine Susceptibility of Cells to EV-D68 Infection.

We performed a haploid genetic screen (17, 18) by infecting mutagenized human HAP1 cells with the EV-D68 prototype strain Fermon CA62-1. The screen identified nine genes involved in Sia biology (Fig. 1 A and B and Fig. S1), seven of which were in the top 10. Among these nine hits are genes involved in the biosynthesis [UDP-GlcNAc-2-epimerase/ManAc kinase (GNE) and N-acetylneuraminic acid synthase (NANS)] and activation [cytidine monophosphate N-acetylneuraminic acid synthetase (CMAS)] of N-acetylneuraminic acid, the predominant form of Sia in humans. Other hits include transporters that transfer the activated sugars CMP-Sia and UDP-galactose [solute carrier family 35 member A1 and A2 (SLC35A1 and SLC35A2)] from the cytosol to the Golgi apparatus and four glycosyltransferases responsible for conjugation of N-acetylglucosamine (GlcNAc) to mannose residues in N-linked glycans [mannoside acetylglucosaminyltransferase 5 (MGAT5)], galactose [beta-1,4-galactosyltransferase 1 (B4GALT1)], and Sia, either via α2,3 linkage [ST3 beta-galactoside alpha-2,3-sialyltransferase 4 (ST3GAL4)] or α2,6 linkage [ST6 beta-galactosamide alpha-2,6-sialyltranferase 1 (ST6GAL1)]. Together, these data provide insights into the identity and composition of the EV-D68 receptor, pointing to an important role of α2,6- and α2,3-linked Sia on N-linked glycans in infection.

Fig. 1.

A haploid genetic screen for EV-D68 identifies genes involved in synthesis of sialylated glycans. (A) An overview of the different steps in synthesis of Sia and sialylated N-linked glycans. Significant hits related to Sia biology are shown in green. Gal, galactose; GlcNAc, N-acetylglucosamine; Man, mannose; Sia, sialic acid. (B) RefSeq gene structures of CMAS and SLC35A1 are shown with the transcriptional orientation pointing from left to right. Gene-trap insertions predicted to disrupt gene function (intronic insertions in sense orientation and insertions mapping to exons) are illustrated for HAP1 cells selected with EV-D68 (selected) and a cell population of similar complexity that had not been selected with EV-D68 (unselected). The FDR-corrected P values for the enrichment of disruptive gene-trap mutations in the EV-D68–selected population are indicated (Materials and Methods). For insertion plots of the other hits, see Fig. S1.

Fig. S1.

Enrichment of gene-trap insertions in genes related to synthesis of sialylated glycans. RefSeq gene structures of identified genes related to Sia biology are shown with the transcriptional orientation pointing from left to right. Gene-trap insertions predicted to disrupt gene function (intronic insertions in sense orientation and insertions mapping to exons) are illustrated for HAP1 cells selected with EV-D68 (selected) and a cell population of similar complexity that had not been selected with EV-D68 (unselected). The FDR-corrected P values for the enrichment of disruptive gene-trap mutations in the EV-D68–selected population are indicated (Materials and Methods).

To confirm the results of the genetic screen, we used mutant cell lines lacking surface expression of Sia (SLC35A1^KO and CMAS^KO) or having a defect in formation of α2,3- and/or α2,6-linked Sia (ST3GAL4^KO, ST6GAL1^KO, and ST3GAL4/ST6GAL1^DKO). The integrity of these mutant cell lines was confirmed by genetic analysis (Fig. S2 A and B), lectin stainings (Fig. S2C), and infection with Sia-dependent and -independent control viruses. Analysis of the number of infected cells showed that SLC35A1^KO and CMAS^KO cells were highly resistant to influenza A virus (IAV), whereas ST3GAL4/ST6GAL1^DKO cells were partially resistant (Fig. 2A), consistent with the broad Sia specificity of IAV (19). In contrast, coxsackievirus-B3, which does not require Sia, could infect all cell lines (Fig. 2 A and B), whereas equine rhinitis A virus, a picornavirus that requires α2,3-linked Sia (20), could infect ST6GAL1^KO but not ST3GAL4^KO cells (Fig. 2C). Upon characterizing EV-D68 Fermon, we observed that infection was inhibited by NA treatment and almost completely blocked in CMAS^KO, SLC35A1^KO, and ST3GAL4/ST6GAL1^DKO cells, both at high (Fig. 2A and Fig. S2D) and low multiplicity of infection (moi) (Fig. 2B). Likewise, little, if any, production of progeny virus was observed in these mutant cell lines (Fig. 2C).

Fig. S2.

Generation and characterization of Sia-deficient HAP1 cells. (A) Genetic characterization of ST6GAL1-deficient and ST6GAL1/ST3GAL4-deficient HAP1 cells. (Top) Depiction of the ST6GAL1 locus with CRISPR target sites upstream (CRISPRup1) and downstream (CRISPRlo1). (Bottom) Cotransfection of guide RNAs for both target sites leads to Cas9-mediated excision of a 148,889 bp fragment encompassing the entire ST6GAL1 locus, fusing the two CRISPR target sites, as revealed by sequencing in a HAP1 clone. (B) ST3GAL4 was targeted by CRISPR-Cas9 in the ST6GAL1-deficient HAP1 clone generated in A. Sequencing reveals a deletion of 2 bp (highlighted in yellow in wild-type control cells) in an early coding exon of ST3GAL4 in a HAP1 subclone, resulting in a frame shift mutation in ST3GAL4 (the amino acid sequence is indicated above the nucleotide sequence). (C) Lectin staining of HAP1 clones. HAP1 cells treated with NA or HAP1 clones having defects in Sia synthesis were stained with S. nigra lectin (SNA; binds to Sia–α2,6-galactose–GlcNAc) and M. amurensis lectin I (MAL1; binds to Sia–α2,3-galactose–GlcNAc). NA treatment resulted in lower MAL1 staining but did not have a profound effect on SNA staining. SNA staining was reduced in cells deficient for CMAS, SLC35A1, and ST6GAL1, confirming the absence of α2,6-linked Sias. Consistently, MAL1 staining was reduced in cells deficient for CMAS, SLC35A1, and ST3GAL4. Although genetic characterization in B and functional characterization (Fig. 2C; ERAV) confirmed the defect in ST3GAL4 expression in ST3GAL4/ST6GAL1 double-deficient cells, no reduction of MAL1 staining was observed in this clone, possibly due to up-regulation of other sialyltransferases forming α2,3-linked Sia that cannot serve as a functional EV-D68 receptor. (D) Both α2,3-linked Sia and α2,6-linked Sia can support EV-D68 infection. Quantification of immunofluorescence images is shown in Fig. 2A. HAP1 cells treated with NA or HAP1 clones having defects in Sia synthesis were infected with EV-D68 Fermon at an moi of 10. Infected cells were visualized by staining with capsid-specific antiserum, imaged by confocal microscopy, and counted. The mean ± SEM of three biological replicates is shown. P values compared with WT were calculated by one-way ANOVA of log-transformed data with Bonferroni correction. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. (E) α2,3-linked and α2,6-linked Sias are heterogeneously expressed in cultured cells. HAP1 cells were fixed either 1 d or 3 d after seeding and stained with S. nigra lectin and M. amurensis lectin I. Patches of cells staining predominantly with SNA or with MAL1 were observed both after 1 d and 3 d of growth, indicating that the differential expression of α2,3- or α2,6-linked Sia between cells is stable over time.

Fig. 2.

Both α2,3- and α2,6-linked Sia can support EV-D68 infection. (A) HAP1 cells treated with NA or HAP1 clones having defects in Sia synthesis were challenged with EV-D68 (Fermon), CV-B3, or IAV-GFP at an moi of 10. Infected cells were visualized by immunostaining and fluorescence microscopy. (B) Cells were infected at an moi of 0.1, and the percentage of infected cells was determined by immunostaining and counting. The mean ± SEM of four biological replicates is shown. (C) Yields of infectious virus after a single cycle of virus production. ERAV, equine rhinitis A virus; EV-D70, enterovirus-D70; EV-D94, enterovirus-D94. The mean ± SEM of three biological replicates is shown. In panels B and C, P values compared with WT were calculated by one-way ANOVA of log-transformed data with Bonferroni correction. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. (D) ST3GAL4/ST6GAL1^DKO cells transfected with plasmids containing ST3GAL4 or ST6GAL1 cDNA were exposed to EV-D68, and yields of infectious virus were measured after a single replication cycle. (E) ST6GAL1^KO or ST3GAL4/ST6GAL1^DKO cells (ST3/ST6^DKO) transduced with retrovirus expressing ST3GAL4 cDNA were exposed to EV-D68 and stained with antibody against the viral capsid.

EV-D68 Can Use Both α2,6- and α2,3-Linked Sia to Infect Cells.

Specificity for α2,3- or α2,6-linked Sia can greatly affect tissue tropism of respiratory viruses, as the Sia abundance varies between the upper (mainly α2,6-linked) and lower (mainly α2,3-linked) respiratory tract (21, 22). Identification of ST3GAL4 and ST6GAL1 suggested that both α2,6- and α2,3-linked Sia are used for infection. Indeed, ST6GAL1^KO cells were less susceptible to EV-D68, whereas ST3GAL4^KO cells were equally susceptible as wild-type cells (Fig. 2 A–C), suggesting a preference of EV-D68 for α2,6-linked Sia. However, EV-D68 can also use α2,3-linked Sia as a receptor, as shown by the observation that ST3GAL4/ST6GAL1^DKO cells were more resistant to infection than ST6GAL1^KO cells (Fig. 2 A–C). Consistently, ST3GAL4/ST6GAL1^DKO cells could be rendered more susceptible to EV-D68 by transfection of plasmid containing ST3GAL4 cDNA (and also ST6GAL1), but the effect was small due to the low transfection efficiency in HAP1 cells (Fig. 2D). Upon retroviral transduction with ST3GAL4 cDNA, however, high infection efficiency was observed (Fig. 2E).

Identification of ST3GAL4 in the genetic screen is likely due to heterogeneous expression of α2,3- and α2,6-linked Sia, which has been described in human airway epithelial cultures but also occurs in cultured cells, including HAP1 (Fig. S2E). Hence, knockout of ST3GAL4 in cells already expressing α2,6-linked Sia at reduced levels likely conferred resistance to infection. In summary, our data show that both α2,6- and α2,3-linked Sia can be used for infection by EV-D68.

Other EV-D Members Display a Similar Sia Preference Profile.

We also investigated the Sia dependency of EV-D70 and EV-D94, two other members of the EV-D species. EV-D70 causes outbreaks of hemorrhagic conjunctivitis, which are often associated with neurological disorders (23). EV-D94 has been associated with acute flaccid paralysis (24), but information on the clinical relevance of this virus is scarce. EV-D70 was reported to be NA-sensitive (25), whereas any role of Sia in EV-D94 infection is unknown. Like EV-D68 Fermon, EV-D70 and EV-D94 could not replicate in CMAS^KO and SLC35A1^KO cells. Also, ST3GAL4/ST6GAL1^DKO cells were less susceptible to infection than ST6GAL1^KO cells (although differences for EV-D70 were less pronounced), suggesting that EV-D70 and EV-D94 can use both α2,6- and α2,3-linked Sia as the receptor (Fig. 2C). Altogether, these data suggest conserved Sia use between EV-D members.

Identification of EV-D68 Strains That Can Infect Cells in a Sia-Independent Manner.

EV-D68 Fermon was isolated more than 50 y ago and genetically differs from currently circulating strains. Therefore, we investigated the Sia dependence of six recent EV-D68 strains—belonging to the three different EV-D68 clades—isolated from patients with respiratory infections in 2009 and 2010 in the Netherlands (5). To minimize the emergence of genetic variants, these isolates were subjected to a limited number of passages in cell culture before they were analyzed for their ability to infect different HAP1 cell lines. Like EV-D68 Fermon, three clinical strains (670, 2042, and 2284) did not cause cytopathic effect (CPE) in Sia-deficient cells (Fig. 3A) and were sensitive to NA treatment, as shown by analysis of virus production in a single cycle (Fig. 3B) or multiple cycles of replication (Fig. S3 A and B). These strains, like Fermon, exhibited some virus production in ST6GAL1^KO cells, despite the lack of CPE (Fig. S3A). Strikingly, the other three strains (947, 1348, and 742) were able to replicate in Sia-deficient and NA-treated HAP1 cells (Fig. 3 A and B). Sia independence of these strains was not specific for HAP1 cells, as these viruses also efficiently infected NA-treated A549 and HeLa-R19 cells (Fig. S3C). In summary, these data demonstrate that several recent EV-D68 strains strongly depend on Sia, whereas other strains can infect cells in a Sia-independent manner, pointing toward the use of a nonsialylated receptor.

Fig. 3.

Identification of EV-D68 strains that can infect cells independently of Sia. (A) HAP1 clones were infected with different EV-D68 strains and stained with crystal violet. (B) HAP1 cells treated with NA or CMAS^KO cells were infected with different EV-D68 strains, and yields of infectious virus were measured after a single cycle of replication. Virus input levels (T = 0) are indicated by a dashed line. ND, not detectable. The mean ± SEM of three biological replicates is shown.

Fig. S3.

Replication and virus production of recent EV-D68 isolates on Sia-deficient cells. (A) Virus production of EV-D68 isolates on Sia-deficient HAP1 clones. HAP1 cells treated with NA and HAP1 clones lacking Sia (CMAS^KO) or having a defect in α2,3-linked Sia (ST3GAL4^KO) or α2,6-linked Sia (ST6GAL1^KO) were infected with different EV-D68 strains or coxsackievirus B3 (CV-B3) at an moi of 0.01. Cells were incubated for 3 d, followed by measurement of virus titers. ND, not detectable. The mean ± SEM of three biological replicates is shown. (B) Comparison of Sia-dependent and -independent EV-D68 isolates in a multicycle virus production assay. HAP1 cells treated with NA or HAP1 cells deficient for CMAS (CMAS^KO) were infected with different EV-D68 strains or CV-B3 at an moi of 0.01 and incubated for ∼3 d (EV-D68) or 1 d (CV-B3), followed by measurement of virus titers. ND, not detectable. The mean ± SEM of three biological replicates is shown. (C) Sia dependence of EV-D68 isolates on other cell lines. A549 (human lung carcinoma) and HeLa-R19 cells treated with NA were infected with different EV-D68 strains at an moi of 0.05 and incubated for 24 h, followed by measurement of virus yields. Virus input levels (T = 0) are indicated by a dashed line. ND, not detectable. The mean ± SEM of three biological replicates is shown.

Sia-Independent EV-D68 Strains Retain Sia-Binding Capacity.

A recent study showed that, in vitro, Sia-containing trisaccharides can bind to EV-D68, where the floor of the canyon would be in the major group rhinoviruses or polioviruses (26) (Fig. 4). To gain insight into the residues that allow Sia-independent infection and their location with respect to the Sia-binding site, we sequenced the capsid regions of the different EV-D68 strains. Amino acid sequence comparison showed that the genetically related Sia-dependent (670) and -independent (742 and 1348) strains in clade A differed at 10 positions, whereas the Sia-dependent (2042) and -independent (947) strains in clade B differed at seven positions (Fig. 4 and Tables S1 and S2). Although residues that differed between Sia-dependent and -independent strains in clade A show little overlap with those that were altered in clade B, some of the changed residues on the viral surface are near the Sia-binding site (26) (Fig. 4).

Fig. 4.

A map of the EV-D68 surface residues. Surface representation of the complete EV-D68 capsid structure (12) (Left) and a close-up showing the Sia-binding site within the canyon (Right). The black triangle shows an icosahedral asymmetric unit of the EV-D68 Fermon structure. Residues are colored by radial distance (Å) to the virus center. The black dashed lines outline the canyon in the human rhinovirus 14 structure. Variable residues determining Sia (in)dependence of strains within clade A or clade B are highlighted by black and white contours, respectively. Note that residue 270 in VP1 (black and white contour) is variable both within clade A and clade B. Amino acids were numbered according to Liu et al., 2015 (12).

Table S1.

Capsid residue differences between EV-D68 clade A strains

Protein	Location	Amino acid residue
		670, Sia dependent	742, Sia independent	1348, Sia independent
VP2	138	T	T	N
	142	R	W	G
	247	A	A	T
VP3	47	V	V	I
	238	Y	H	H
VP1	85	Q	R	Q
	133*	T	R	T
	155*	K	E	E
	205*	S	S	N
	269*	R	K	K

Genomic sequences used for this comparison are 4311000670_tMK3RD3 (GenBank accession no. KT231906), 4311000742_tMK3RD3 (GenBank accession no. KT231908), and 4310901348_RD3 (GenBank accession no. KT231900).

^*These residues correspond to VP1 residues 134, 156, 206, and 270, respectively, in clade B strains and EV-D68 Fermon.

Table S2.

Capsid residue differences between EV-D68 clade B strains

Protein	Location	Amino acid residue
		2042, Sia dependent	947, Sia independent
VP2	29	V	A
	116	K	R
VP3	52	L	M
	59	K	E
VP1	157	R	K
	270	R	K
	271	E	K

Genomic sequences used for this comparison are 4310900947_RD3 (GenBank accession no. KT231898) and 4310902042_RD3 (GenBank accession no. KT231902).

To investigate whether the amino acid substitutions that established an alternative receptor-binding site might have affected the Sia-binding capacity of these strains, we performed hemagglutination experiments with blood from nine different human donors. All EV-D68 isolates agglutinated human erythrocytes, although two Sia-independent strains (1348 and 742) agglutinated erythrocytes from only one donor (Table 1). Pretreatment of erythrocytes with NA prevented hemagglutination by EV-D68 strains but not by echovirus-7, which agglutinates erythrocytes by binding to its protein receptor, decay-accelerating factor (27). Remarkably, blood from three donors was not agglutinated by any of the EV-D68 strains, whereas hemagglutination titers of IAV were similar for all donors, indicating equal Sia expression levels. This variability suggests that EV-D68 does not merely bind any sialylated glycan but has a preference for specific sialylated glycan structures that are differentially expressed between individuals. No clear correlation between EV-D68 hemagglutination and ABO blood groups was observed. In summary, these data indicate that Sia-independent strains have retained their Sia-binding capacity, albeit two strains (742 and 1348) seem to have a reduced affinity for Sia.

Table 1.

Hemagglutination titers (Log₂) of EV-D68 strains on human erythrocytes from nine different donors

Donor	1	2	3	4	5	6	7	8	9	1 + NA	2 + NA
EV-D68 2042	2	2	0	2	4	0	2	0	6	0	0
EV-D68 947	3	2	0	3	4	0	2	0	6	0	0
EV-D68 2284	2	2	0	3	4	0	2	0	5	0	0
EV-D68 670	3	2	0	3	3	0	2	0	7	0	0
EV-D68 742	0	0	0	0	0	0	0	0	3	0	0
EV-D68 1348	1	0	0	0	0	0	0	0	3	0	0
IAV PR8	9	9	9	10	10	9	9	9	10	0	0
IAV WSN	8	8	8	9	8	8	7	8	8	0	0
Echovirus 7	8	7	ND	ND	ND	ND	ND	ND	ND	8	9

ND, not determined.

Discussion

In this study, we provided important insights into the identity/nature of the EV-D68 receptor. Using a genome-wide haploid screen, we identified genes involved in biosynthesis (GNE and NANS), activation (CMAS), transport (SLC35A1), and conjugation of Sia to glycans (ST3GAL4 and ST6GAL1) as factors required for EV-D68 infection. Using knockout cell lines and gene reconstitution, we have shown that EV-D68 can use both α2,6- and α2,3-linked Sia as a receptor to infect cells. This finding extends recent observations that both α2,6- and α2,3-linked Sia-containing trisaccharides can bind to the EV-D68 capsid and initiate virion uncoating in vitro (26). The observation that EV-D68 can use not only α2,6-linked Sia as a receptor but also α2,3-linked Sia, which resides mainly in the lower respiratory tract, may provide an explanation for its ability to cause severe lower respiratory tract infections.

Importantly, the screen did not identify a specific protein receptor, suggesting that EV-D68 can use multiple redundant receptors, given these are glycosylated with a suitable sialylated glycan. Our screen also provided insights into the preference of EV-D68 for specific sialylated glycans. Identification of glycosyltransferases responsible for conjugation of galactose (B4GALT1) and GlcNAc (MGAT5) pointed to the importance of Sia–galactose–GlcNAc chains, consistent with the substrate specificity of ST3GAL4 and ST6GAL1. Furthermore, the identification of MGAT5, which forms GlcNAc–β1,6-Man linkages, suggested that EV-D68 specifically recognizes N-linked glycans containing a β1,6-linked antenna. It should be noted that Sia–galactose–GlcNAc, although mainly expressed on N-linked glycans, also occurs on O-linked glycans and glycolipids and that we observed that HEK293S cells, which lack complex N-linked glycans (19), are susceptible to EV-D68. Further proof that EV-D68 does not merely bind any sialylated glycan but has a preference for specific glycans stems from our observation that erythrocytes of several donors could be agglutinated by IAV but not by EV-D68. More research is required to explore the glycan spectrum that can be bound by EV-D68.

Upon characterizing recent EV-D68 isolates, we identified strains that are able to infect Sia-deficient cells, implying that these viruses can use an alternative entry receptor. Genetic comparison of Sia-dependent and -independent EV-D68 strains within clades A and B revealed little overlap of residues determining Sia independence but pointed toward residues near the Sia-binding site as possible determinants for Sia (in)dependence. A similar scenario was described for Sia-dependent and -independent rotavirus strains, where only a few amino acid changes in the Sia-binding site could cause a receptor switch to a nonsialylated glycan (28). It remains to be investigated whether a single amino acid substitution or a combination thereof is required for the observed Sia-independent phenotype. Better understanding of the (combinations of) residues that facilitate Sia independence could ultimately allow prediction of receptor requirement based on sequence alignments.

It has been shown in vitro that Sia binds to the EV-D68 canyon at a unique site, compared with the glycan-binding sites in other picornaviruses (Fig. S4). In EV-D68, binding of Sia induces virion destabilization and pocket factor release, which is the first step in the uncoating process (26). We identified several EV-D68 strains that can use a nonsialylated receptor while retaining a Sia-binding capacity, albeit with different affinities. This indicates that the formation of an alternative receptor-binding site does not necessarily result in loss of the Sia-binding site and points to the possible existence of a dual receptor mechanism, where either Sia or a nonsialylated receptor can trigger similar structural conformational changes. However, it is unclear whether binding of Sia to the capsid of Sia-independent viruses still results in virion destabilization and pocket factor release as described for EV-D68 Fermon. Furthermore, more research is warranted to determine whether this alternative receptor is a protein or a sugar that lacks a terminal Sia moiety.

Fig. S4.

Comparison of carbohydrate receptor binding sites on picornaviruses. All known picornavirus–carbohydrate receptor structures were superimposed. Two adjacent protomers of the EV-D68 capsid are represented as Cα backbones. The triangle outlines an icosahedral asymmetric unit of the virus capsid. VP1, VP2, and VP3 are colored blue, green, and red. Carbohydrate receptors are shown as sticks. Cyan, Neu5Ac [coxsackievirus A24 variant, Protein Data Bank (PDB) ID code 4Q4X]; magenta, 3′-sialyllactose (equine rhinitis A virus, PDB ID code 2XBO); yellow, 6'-sialyllactose (EV-D68, PDB ID code 5BNN); dark green, heparan sulfate (foot-and-mouth disease virus, PDB ID code 1QQP).

Although it remains to be established whether Sia-independent strains circulate in the human population, the occurrence of strains that use an alternative receptor could have an impact on tissue tropism and pathogenesis of EV-D68. Also, application of the sialidase DAS-181 (Fludase) (29), an investigational drug against influenza virus that was shown to inhibit EV-D68 (30), may be ineffective against Sia-independent EV-D68 strains. Hence, detailed insight into the interactions of EV-D68 with its receptor(s) is required to understand viral pathogenesis and to develop effective antiviral treatment.

Materials and Methods

Cells and Viruses.

Information on viruses and cells used in this study is described in SI Materials and Methods.

Haploid Genetic Screen with EV-D68.

HAP1 cells were gene-trap mutagenized as described previously (31). Following expansion, 10⁸ mutagenized cells were exposed to EV-D68 Fermon (moi 3). After selection, surviving cells were expanded and used for genomic DNA isolation. Insertion sites identified in cells selected with EV-D68 (yielding 414,290 unique gene-trap insertions mapped to genes) and a population of matched control cells of comparable complexity (495,679 unique gene-trap insertions mapped to genes) were aligned to the human genome not filtering for close reads (31). Subsequently, disruptive insertion sites (in sense orientation of the affected gene or mapping to exons) in significantly identified genes were compared in the two cell populations, and P values for enrichment were calculated using a Fisher’s exact test as described previously (31). Disruptive insertion sites in virus-selected and control cells were plotted onto the RefSeq gene bodies for the following transcripts: NM_001497 (B4GALT1), NM_018686 (CMAS), NM_001128227 (GNE), NM_002410 (MGAT5), NM_018946 (NANS), NM_006416 (SLC35A1), NM_005660 (SLC35A2), NM_006278 (ST3GAL4), and NM_173216.2 (ST6GAL1).

Generation of Knockout Cells.

ST3GAL4^KO and SLC35A1^KO HAP1 cells have been described (31). CMAS^KO cells were obtained from Haplogen GmbH. The CRISPR-Cas9 system was used to generate ST6GAL1^KO cells. The entire ST6GAL1 locus was excised (Fig. S2A), and subclones were analyzed by genotyping (Table S3). ST3GAL4/ST6GAL1^DKO cells were obtained by deleting an exonic region in ST3GAL4 from ST6GAL1^KO cells using CRISPR-Cas9.

Table S3.

Oligonucleotide primers used in this study

Primer use	Sequence, 5′–3′
Genotyping HAP1 ST6GAL1KO: excision of ST6GAL1	ATTCTAGATAAGAGCAACACGAGTCTTC
Genotyping HAP1 ST6GAL1KO: excision of ST6GAL1	TGAGAGTGGAAAGCGGTGTTAGAGC
Genotyping HAP1 ST3GAL4/ST6GAL1KO: frame shift in ST3GAL4	GGTTCCAAGTGGAACTTAACT
Genotyping HAP1 ST3GAL4/ST6GAL1KO: frame shift in ST3GAL4	CTTCTCTTCCCTCCCTGGCTT
EV-D68 cDNA synthesis	TTTTTTTTTTTTTTTTTTTTGG
Sequencing complete genome of clade B isolates	TGCCAGTGGAATGAATCTTGC
Sequencing complete genome of clade B isolates	ACATTAAAATGCGGTGCGTT
Sequencing complete genome of clade B isolates	TCAGAGGATTCACTGGGGAC
Sequencing complete genome of clade B isolates	CCTGAGCTCCCATTATTAAAATCT
Sequencing complete genome of clade B isolates	TAGTAACAACACATACATGGGTCTTC
Sequencing complete genome of clade B isolates	CATAGCTGCCCAACAACAGA
Sequencing complete genome of clade B isolates	GCGTACATACATCAAAATTCCC
Sequencing complete genome of clade B isolates	TTGTCAAATCAATTGCAAAAGG
Sequencing complete genome of clade B isolates	AAGTTGAAAAAGGGAAATCACG
Sequencing complete genome of clade B isolates	GCCCACCCGTGAAGGTG
Sequencing complete genome of clade B isolates	CGGAGGCTTTTGACTTTCTGCT
Forward primer for amplification of capsid region from clade A and C isolates for sequencing	TATCCCGGGTTCTTAAAACAGCCTTGGGGT
Reverse primer for amplification of capsid region from clade A and C isolates for sequencing	GATATCAGTGAAAGCTACAAT
Sequencing capsid region of clade A and C isolates	ATTGTTACCATTTAGCTTGTCAAAT
Sequencing capsid region of clade A and C isolates	GTGATATGAGATCCATTTGTGGC
Sequencing capsid region of clade A and C isolates	CATGAAAGGTGAAGAAGGAGGGA
Sequencing capsid region of clade A and C isolates	CCTCCAGGTGGGTCATG
Sequencing capsid region of clade A and C isolates	CCCCCAGCCAGAATGAC

Infectivity Assays.

Cells were infected with virus for 1 h. After incubation for the indicated period, virus titers were determined by end-point dilution. Crystal violet staining was performed at 3 d postinfection. Where indicated, cells were pretreated with NA from Clostridium perfringens (NEB) or from Arthrobacter ureafaciens (Roche) in serum-free medium for 30 min.

Immunofluorescence Assays.

Paraformaldehyde-fixed cells were stained using rabbit anti-capsid serum against EV-D68 Fermon (produced in house; 1:1,000) or a mouse monoclonal antibody against CV-B3 protein 3A (1:100) (32). For characterization with lectins, cells were stained with fluorescein-labeled Sambucus nigra lectin (Vector Laboratories; 1:1,000) and biotinylated Maackia amurensis lectin I (Vector Laboratories; 1:500). Cells were examined by confocal microscopy (Leica SPE-II) or standard fluorescence microscopy (EVOS FL cell imaging system). The number of nuclei was quantified using ImageJ, and the number of infected cells was quantified visually.

Isolation and Sequencing of EV-D68 Strains from Clinical Specimen.

Monolayers of tertiary monkey kidney cells (tMKs) or human rhabdomyosarcoma (RD) cells were incubated with 250 μL EV-D68–positive clinical material (mixed nose and throat swabs) derived from patients with influenza-like illness or acute respiratory infection (5) and incubated at 34 °C until CPE was observed. EV-D68 strains 4311000670 (clade A; further referred to as 670) and 4311000742 (clade A; 742) were isolated on tMK cells, whereas strains 4310900947 (clade B; 947), 4310901348 (clade A; 1348), 4310902042 (clade B; 2042), and 4310902284 (clade C; 2284) were isolated on RD cells. Viruses were harvested by one freeze-thawing cycle at –80 °C and clarified by centrifugation. Subsequent passages of virus were done on tMK or RD cell monolayers. The full genome of the virus isolates was sequenced from passage 4 for strains 670 and 742 and from passage 2 of the remaining four strains. All infection experiments described in this study were performed with viruses that had undergone one or two more rounds of passage on RD cells. From these viruses, the 5′UTR and capsid region was sequenced. The latter sequences were used for amino acid comparisons shown in Tables S1 and S2. Detailed information on virus passages, sequences, and GenBank accession numbers is described in SI Materials and Methods. As required by Dutch legislation, surveillance studies have to be registered in the Personal Data Protection Act Register of the Personal Data Protection Commission. The influenza surveillance from which the clinical enterovirus D68 isolates were obtained is registered in this register and no further ethical approval was needed for this virologic study because only anonymized virus isolates were used.

Hemagglutination Assays.

The hemagglutination assay was performed using standard methods. Briefly, twofold dilutions of the virus stocks were incubated with an erythrocyte suspension for 16 h at 4 °C.

SI Materials and Methods

Cells and Viruses.

HAP1 cells (Haplogen GmbH) were cultured in Iscove's Modified Dulbecco's Medium (Lonza) supplemented with 10% (vol/vol) FCS. HeLa-R19, A549, HEK293T (ATCC CRL-3216), and RD cells were cultured in DMEM (Lonza) supplemented with 10% (vol/vol) FCS. All cell lines were tested for mycoplasma contamination. HeLa-R19 cells were obtained from G. Belov (University of Maryland and Virginia-Maryland Regional College of Veterinary Medicine, Blacksburg, VA) (33). RD cells were obtained from the European Collection of Cell Cultures (cat. no. 85111502). A549 cells, EV-D68 (Fermon CA62-1), EV-D70 (Morocco), EV-D94 (E210), and echovirus 7 (Wallace) were obtained from the National Institute of Public Health and the Environment. CV-B3 (Nancy) was obtained by transfecting in vitro-transcribed RNA derived from full-length infectious clone p53CB3/T7 as previously described (34). ERAV (NM11/67) was obtained from David Rowlands and Toby Tuthill (University of Leeds, Leeds, United Kingdom). IAV PR8-Sinai and IAV–PR8-Sinai–GFP were from Adolfo García-Sastre (Department of Microbiology, Mount Sinai School of Medicine, New York, NY).

Generation of Knockout Cells by CRISPR-Cas9.

The CRISPR-Cas9 system was used to generate ST6GAL1^KO cells. For this, haploid HAP1 cells were cotransfected with a pMX vector encoding a GFP–IRES–Blasticidin S-resistance gene and two px330-based constructs targeting a region upstream of the ST6GAL1 locus (5′-GG TCG CTA GCG AGC GGG CTT GGG-3′) and one downstream of ST6GAL1 (5′-GA GAA CCC TAC TAG TCG GGT GGG-3′) to excise the entire locus (Fig. S2A). One day after transfection, cells were selected with Blasticidin S (30 μg/mL, Invivogen) for 24 h, expanded, and subcloned. Individual subclones were analyzed for the excision of the ST6GAL6 locus by genotyping (Table S3). ST3GAL4/ST6GAL1^DKO cells were obtained by mutating ST3GAL4 in ST6GAL1^KO cells using a px330 construct targeting an exonic region (5′-AG CAA GGC CTC TAA GCT CTT TGG-3′) in ST3GAL4 (Fig. S2B). Cells were selected, expanded, subcloned, and genotyped as described above.

cDNA Complementation in ST3GAL4/ST6GAL1^DKO Cells.

Murine leukemia virus particles for reconstitution of ST3GAL4 were made by cotransfection of pBabe-puro containing ST3GAL4 cDNA (described previously; ref. 31) with pCAGGS-VSV-G and pMLV-gag-pol in HEK293T cells. The supernatant containing the retrovirus was collected 3 d posttransfection and used for transduction of HAP1 ST3GAL4/ST6GAL1^DKO cells. One day later, transduced cells were selected with puromycin (1 μg/mL). Reconstitution of ST3GAL4 and ST6GAL1 with plasmid DNA was performed by transfecting ST3GAL4/ST6GAL1^DKO cells with plasmids encoding either sialyltransferase. Cells were infected with EV-D68 Fermon 24 h posttransfection.

Sequencing of Recent EV-D68 Strains Isolated from Clinical Specimen.

The full genome of the virus isolates was sequenced from passage 4 (tMK3, RD1) for strains 670 and 742 (GenBank accession nos. 4311000670_tMK3RD1, KT231905; 4311000742_tMK3RD1, KT231907) and from passage 2 (RD2) of the remaining four strains (GenBank accession nos. 4310900 947_RD2, KT231897; 4310901348_RD2, KT231899; 4310902042_RD2, KT231901; 4310902284_RD2, KT231903) using 22 primer sets (details available on request). cDNA was synthesized using the Thermoscript RT-PCR System (Invitrogen) followed by PCR using Phusion High-Fidelity PCR Master Mix with GC Buffer (Finnzymes). Amplification products were treated with ExoSAP-IT (Affimetrix), and sequencing reactions were performed using BigDye terminator reagent (Life Technologies) and analysis of product on ABI3700 automated sequencer. Derived trace files were trimmed from primer sequences and assembled using Bionumerics software version 6 (Applied- Maths). All infection experiments described in this study were performed with viruses that had undergone one or two more rounds of passaging on RD cells. From these viruses, either the full genome (GenBank accession nos. 4310900947_RD3, KT231898; 4310902042_RD3, KT231902) or the 5′UTR and capsid region was sequenced (GenBank accession nos. 4311000670_tMK3RD3, KT231906; 4311000742_tMK3RD3, KT231908; 4310901348_RD3, KT231900; 4310902284_RD3, KT231904) using the primers listed in Table S3. Importantly, a similar Sia dependency/independency profile was observed with viruses that had undergone one passage less on RD cells.