Enhanced Enterovirus D68 Replication in Neuroblastoma Cells Is Associated with a Cell Culture-Adaptive Amino Acid Substitution in VP1

Enterovirus D68 (EV-D68) causes mild to severe respiratory disease and is associated with acute flaccid myelitis since 2014. Currently, the understanding of the ability of EV-D68 to replicate in the central nervous system (CNS), and whether it is associated with a specific clade of EV-D68 viruses or specific viral factors, is lacking. Comparing different EV-D68 clades did not reveal clade-specific phenotypic characteristics. However, we did show that viruses which acquired a cell culture-adapted amino acid substitution in VP1 (E271K) recognized heparan sulfate as an additional receptor. Recognition of heparan sulfate resulted in an increase in attachment, infection, and replication in neuroblastoma cells compared with viruses without this specific amino acid substitution. The ability of EV-D68 viruses to acquire cell culture-adaptive substitutions which have a large effect in experimental settings emphasizes the need to sequence virus stocks.

EV-D68 belongs to the Enterovirus genus in the Picornaviridae family. It is a nonenveloped, positive-sense, single-stranded RNA virus. Within the capsid, the structural protein VP1 plays an important role in the attachment to host cells, and its gene is the most variable part of the genome and used for genotyping (18). Circulating EV-D68 isolates use cell surface glycoproteins including sialylated glycoproteins or glycolipid as a receptor. These sialylated receptors bind to the canyon in VP1, which leads to conformational change and subsequent dysregulation of stability and thereby initiates uncoating (19). However, recent studies have identified nonsialylated receptors, such as ICAM-5 and heparan sulfate glycosaminoglycans (GAGs), for specific EV-D68 isolates (20,21).
To date, EV-D68 is divided into four clades, from A to D, although the initial division in three clades, A, B, and C, with subclades A1 and A2, is also used. Some researchers use A for A1 and D for A2 (22,23). EV-D68 clade B is subdivided into subclades B1, B2, and B3 (24), and clade D has recently been subdivided into subclades D1 and D2 (25). Multiple clades circulated during the 2014 outbreak, but subclade B1 was the most prevalent and was also associated with neurological complications. Based on this observation, it was initially thought that the ability to invade and replicate in the central nervous system (CNS) was a recently acquired feature and clade specific (2,3,8,10,26). However, since 2016, subclade B3 and to a lesser extent subclade D1 became predominant, and both were associated with neurological complications (16,(27)(28)(29). In addition, recent studies have shown that multiple isolates from different clades are able to infect both neuronal cells and neuroblastoma cell lines (30) and that non-subclade B1 and B3 isolates were able to cause paralysis in mice after intracranial inoculation. Interestingly, isolates from the same clade differed in their ability to cause paralysis in mice (31). Altogether, this suggests that the ability to invade and replicate in the CNS is not a clade-specific feature. Even though several studies have determined the in vitro phenotypic characteristics of EV-D68 isolates in cells of the CNS, viral factors associated with efficient attachment, infection, and replication have not been identified yet. In addition, genetic analysis of EV-D68 isolates associated with phenotypic characteristics in vitro has not been carried out in previous studies.
In this study, we investigated the ability of EV-D68 isolates from different clades to attach and infect human neuroblastoma cells (SK-N-SH) and determined their replication kinetics. Subsequently, we compared viral sequences in order to identify viral factors associated with increased viral replication.

RESULTS
Clinical enterovirus D68 isolates from different clades replicate in human neuroblastoma cells with different efficiency. Growth curves using a multiplicity of infection (MOI) of 0.01 were generated for clinical isolates of clade A (or A1; here we use A), subclade B1, subclade B2 (B2/039 and B2/947), subclade B3, subclade D1 (or A2; here we use D1), and the prototype Fermon strain in SK-N-SH and rhabdomyosarcoma (RD) cells. Growth curves showed that all viruses replicated in SK-N-SH cells, but virus titers of subclades B2/947 and B3 and Fermon were significantly higher than those of clade A and subclades D1, B1, and B2/039 from 24 h postinoculation (hpi) onward (Fig. 1A). Remarkable, large differences between the two subclade B2 isolates were observed in growth kinetics on SK-N-SH cells. The same pattern was observed in RD cells, but the differences among isolates were smaller. Isolates that replicated most efficiently on SK-N-SH cells also did so in RD cells (Fig. 1B). Overall, these data show that EV-D68 isolates from all clades are able to replicate in SK-N-SH and RD cells but do so with different efficiency.
Replication efficiency is associated with percentage of infection. To study whether the replication efficiency was associated with the infection efficiency of the different EV-D68 isolates, the percentage of infection was determined by immunofluorescence using an anti-VP1 antibody after incubation of viruses on SK-N-SH and RD cells with an MOI of 2 for 8 h. Viruses of clade B2/947 and Fermon, and to a lesser extent B3, which replicated most efficiently in SK-N-SH and RD cells, infected higher percentages of cells than did other isolates ( Fig. 2A and B). These results suggest that the efficient replication measured by the growth kinetics was associated with efficient infection of cells.
A lysine is present at position 271 of capsid protein VP1 in clinical isolates that replicate efficiently in SK-N-SH and RD cells. After the identification of amino acid differences between the two clinical isolates of subclade B2, the 13 amino acid positions that differed were verified in the other clinical isolates to identify amino acid substitutions possibly associated with increased infection and replication efficiency. A lysine (K) at position 271 in the VP1 capsid protein was observed only in clade B2/947 and B3 viruses, the two clinical isolates that replicated most efficiently in SK-N-SH and RD cells. All other clinical isolates had a glutamic acid (E) or valine (V) at position 271 in VP1, and the lab-adapted Fermon carried an aspartic acid (D) ( Table 2). At all other positions that differed between the clade B2 isolates, no common differences were observed between the viruses with different replication efficiencies (Table 2). This indicates that the amino acid substitution E271K might be a

determinant for increased infection and replication in SK-N-SH cells in vitro.
A lysine at position 271 in VP1 is acquired during in vitro passaging. The clinical isolates used in this study were passaged in RD cells to produce virus stocks. To investigate whether a lysine at position 271 in VP1 was present in the original isolate of the patient, sequences from clinical specimens, previous passages, and virus stocks used in this study were generated and aligned. In the clinical specimens, a glutamic acid (E) was found at position 271 in VP1 of both subclade B2/947 (21) and subclade B3. Furthermore, sequencing analysis of all virus passages showed the acquisition of a lysine (K) in passage 3 of subclade B2/947 and passage 2 of subclade B3 on RD cells (Table 3). There were no minor variants present at position 271 of sequences derived from clinical specimens and stock B2/947 and B3 viruses using a 20% cutoff (Table 3). This indicates that viruses with high replication and infection efficiency acquired the E271K amino acid substitution during passaging in RD cells.
Viruses with the E271K substitution in VP1 attached more efficiently to SK-N-SH and RD cells. Since the amino acid at position 271 of EV-D68 resides within the canyon, close to the sialic acid binding site (19), we investigated whether EV-D68 isolates with the specific E271K amino acid substitution attached to higher percentages of SK-N-SH and RD cells. We observed that the subclade B2/947 and subclade B3 isolates which acquired the E271K substitution attached to higher percentages of SK-N-SH and RD cells than did all other isolates. Fermon attached to lower percentages of SK-N-SH and RD cells despite efficient replication and infection ( Fig. 3A and B).
Overall, these observations suggest that the E271K substitution is associated with increased virus attachment in vitro. Viruses with E271K substitution in VP1 recognize sialic acid and heparan sulfate. A previous study has shown that among several EV-D68 strains, subclade B2/947 can recognize heparan sulfate as an additional receptor to sialic acid (21). To investigate whether our viruses with an E271K substitution were able to recognize both sialic acid and heparan sulfate, we determined the ability of these viruses to attach to cells that were pretreated with Arthrobacter ureafaciens neuraminidase and heparinase III to remove sialic acids or heparan sulfate, respectively. Subclade B2/947 and B3 isolates carrying the E271K substitution, and the subclade B2/039 isolate (271E) as control, were included in these analyses. Removal of sialic acids and heparan sulfate was confirmed by measuring mean fluorescent intensity (MFI) with Maackia amurensis lectin (MAL) for ␣(2,3)-linked sialic acid and Sambucus nigra lectin (SNA) ␣(2,6)-linked sialic acid, and with an antibody that recognizes heparan sulfate ( Fig. 4A and D). The percentage of cells to which the virus attached was significantly decreased for all viruses in both SK-N-SH and RD cells after neuraminidase treatment ( Fig. 4B and C). In contrast, heparinase III treatment resulted in decreased attachment of only viruses with the E271K substitution ( Fig. 4E and F). Overall, these results showed that the viruses with the E271K substitution recognize both sialic acid and heparan sulfate on the cellular surface.
Attachment, internalization, and replication of viruses with E271K substitution in VP1 depend in part on heparan sulfate recognition. To investigate if heparan sulfate is required for attachment, internalization, and replication of viruses with an E271K substitution, heparan sulfate was removed on SK-N-SH and RD cells by either  heparinase III or NaClO 3 , which prevents cell surface sulfation (21). Treated and untreated cells were inoculated with an MOI of 1 of B2/947, B3, and B2/039. After 1 hpi at 37°C, the amount of attached and internalized viral RNA of B2/947 and B3 was smaller in the heparinase III-or NaClO 3 -treated cells than in untreated cells by qPCR. Viral RNA for B2/039 was not reduced in the heparinase-or NaClO 3 -treated cells, with the exception of a minimal but significant reduction of viral RNA in SK-N-SH cells treated with NaClO 3 but not in RD cells ( Fig. 5A and B). Virus replication of both B2/947 and B3 was significantly decreased in SK-N-SH and RD cells treated with heparinase III or NaClO 3 , while replication levels of B2/039 were similar in treated and untreated cells ( Fig. 5C and D; see also Fig. S1 in the supplemental material). Together, these results indicate that the recognition of heparan sulfate is important for facilitating virus attachment, internalization, and replication.

DISCUSSION
Here, we report that EV-D68 isolates from (sub)clades A, B1, B2, B3, and D1 replicated in human neuroblastoma SK-N-SH cells; however, large differences were observed in replication efficiency. Subsequent genotypic analysis revealed that viruses replicating most efficiently had a specific amino acid substitution in VP1 (E271K). This substitution was acquired during cell culture propagation and was associated with the recognition of heparan sulfate. Furthermore, attachment, internalization, and replication depended in part on heparan sulfate recognition for viruses with the E271K substitution in VP1. These findings suggest that efficient replication of EV-D68 isolates in vitro in human neuroblastoma cells is not a clade-specific feature but at least in part associated with the usage of heparan sulfate as an additional receptor.
The amino acid substitution at position 271 of VP1 (E271K) acquired during in vitro passaging in 2 out of the 6 viruses included in our study was associated with the recognition of heparan sulfate as an additional receptor (21), which had a large impact on the phenotypic characteristics of EV-D68 viruses. Heparan sulfate is abundantly expressed on RD cells (32) and associated with increased virus attachment, infection efficiency, and virus replication in cells in vitro. The ability of EV-D68 to replicate in mouse neuroblastoma cells and human induced pluripotent stem cell motor neurons were independent of sialic acid recognition, although the usage of heparan sulfate was not investigated (30,33). In our study, unfortunately, viruses from earlier passages without E271K in VP1 were not available and an infectious clone of the clinical isolate B2/947 with the original VP1 amino acid E271 could not be recovered (21), which made a direct comparison of viruses with and without E271K in VP1 not possible. The lysine at position 271 of VP1 is positively charged, which contributes to the formation of a basic patch that interacts with negatively charged heparan sulfate proteoglycan (21). This mechanism has been described previously for other picornaviruses which acquired the recognition of heparan sulfate during passaging of viruses in vitro (34)(35)(36).
It was initially thought that the neurotropic potential of EV-D68 was a feature recently acquired and was associated with subclade B1. However, epidemiological and clinical data have shown that different subclades, namely, B3 and D1, besides subclade B1 have also been associated with AFM cases after 2014 (16,28,29). In addition, most in vitro studies, including ours, and in vivo studies do not reveal phenotypic difference between clades or recent and older EV-D68 isolates (30,31,37). In vivo, EV-D68 isolates from multiple clades (e.g., clade A and subclade B1) were able to cause paralysis after intracranial inoculation (31). In vitro, we and others showed that EV-D68 isolates from multiple clades were able to replicate in neuroblastoma cells (37). Furthermore, we did not detect phenotypic differences between viruses isolated before and after 2014 independent of the clade, which correlates with the study by Rosenfeld et al. (30), but not with a study performed by Brown et al. (37). Unfortunately, virus stocks were not compared with clinical isolates in these studies. All together, these data suggest that the ability to replicate in cells of the CNS is not a clade-specific feature. The role of heparan sulfate as an additional receptor in the pathogenesis in vivo remains unclear. Although the presence of the E271K substitution in human EV-D68 isolates has not been studied, we found that only a very few EV-D68 VP1 sequences available in GenBank have a lysine at position 271 of VP1 (by 15 December 2019, 17 out of 1,737 [Ͻ1%]), and they fall in a wide range of old and current clades. However, of most sequences the source, clinical material or virus isolate, and passage history are not known. Interestingly, acquisition of heparan sulfate recognition in an immunocompromised patient with enterovirus 71 (EV-71) was associated with systemic spread, including the CNS. In this patient a mutation in the VP1 responsible for heparan sulfate recognition was detected only in extrarespiratory samples (38). Since heparan sulfate is not expressed on the apical side of respiratory epithelial cells, but abundantly in cells outside the respiratory tract, such as cells of the CNS and muscle cells (38,39), in vivo acquisition of E271K might-just like EV-71-occur outside the respiratory tract and would therefore hardly be found in respiratory samples. Sequence analysis of respiratory and extrarespiratory samples from the same patient should reveal if heparan sulfate recognition is important for the pathogenesis of EV-D68 in vivo.
It remains unclear why there has been an increase of AFM cases associated with EV-D68 since 2014, since epidemiological and serological studies suggest that EV-D68 has been circulating throughout the population for decades (40,41). Currently, the pathogenesis of neurological disease associated with EV-D68 infection is not fully understood. Either a direct effect of virus entry and replication in the nervous system or an indirect effect of systemic cytokines could contribute. However, the detection of EV-D68 or virus-specific antibodies in the cerebrospinal fluid suggests that virus enters and replicates in the CNS in at least some of the patients (42,43). As there have been more cases with severe respiratory disease caused by EV-D68 since 2014 (9,44), it might be that recent EV-D68 viruses replicate to higher titers in the respiratory tract, thereby increasing the risk of systemic spread including that to the CNS. However, understanding the exact mechanism of systemic spread of EV-D68 in vivo requires more in-depth pathogenesis studies.
Taken together, we demonstrate that the replication of EV-D68 isolates in the neuroblastoma cell line SK-N-SH is not a clade-specific phenotype of EV-D68. However, large phenotypic differences that we did observe in vitro could be linked to the substitution E271K in VP1 leading to the recognition of heparan sulfate as an additional receptor, which resulted in increased attachment, internalization, and replication. Therefore, we recommend sequencing of virus stocks in order to obtain viruses that resemble clinical isolates in order to study phenotypic characteristics of EV-D68 isolates in vitro and in vivo.
Viruses. Enterovirus D68 viruses included in this study were isolated from clinical specimens at the National Institute of Public Health and the Environment (RIVM), Bilthoven, The Netherlands. The viruses were initially isolated on RD cells at 33°C at RIVM from patients with respiratory disease caused by EV-D68 infection. Virus stocks for the in vitro studies were grown at the Viroscience laboratory, Erasmus MC (EMC), Rotterdam, The Netherlands, in RD cells (ATCC) at 33°C in 5% CO 2 msphere.asm.org 8 eluted in 50 l. For viral RNA quantification, a real-time TaqMan RT-PCR assay was performed using the Applied Biosystems 7500 real-time PCR system (Thermo Fisher Scientific). The experiments were carried out by adding forward and reverse EV-D68 primer (75 pmol/l) and the probes (10 pmol/l). The following sequences including EV-D68 specific primer and probe were used: forward primer 5=-TGTTCC CACGGTTGAAAACAA-3=, reverse primer 5=-TGTCTAGCGTCTCATGGTTTTCAC-3=, probe 1 5=-TCCGCTATAG TACTTCG-3=, and probe 2 5=-ACCGCTATAGTACTTCG-3=. The following reaction conditions were applied for all PCR experiments: 5 min at 50°C and 20 s at 95°C, followed by 45 cycles at 95°C for 3 s, and 60°C for 31 s. Statistical analyses. Statistical analyses were performed using GraphPad Prism 6.0 software (La Jolla, CA, USA). Specific tests are described in the figure legends. P values of Յ0.05 were considered significant. All data were expressed as means Ϯ standard deviations (SDs). Experiments were performed at least in biological triplicates and technical duplicates.
Data availability. The viruses included in this study are available under accession numbers MN954536, MN954537, MN954538, MN954539, MN954540, and MN954541.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, DOCX file, 0.2 MB.