VP1 is the primary determinant of neuropathogenesis in a mouse model of enterovirus D68 acute flaccid myelitis

ABSTRACT

Enterovirus D68 (EV-D68) is an emerging pathogen that can cause severe respiratory and neurologic disease [acute flaccid myelitis (AFM)]. Intramuscular (IM) injection of neonatal Swiss Webster (SW) mice with US/IL/14-18952 (IL52), a clinical isolate from the 2014 EV-D68 epidemic, results in many of the pathogenic features of human AFM, including viral infection of the spinal cord, death of motor neurons, and resultant progressive paralysis. In distinction, CA/14-4231 (CA4231), another clinical isolate from the 2014 EV-D68 outbreak, does not cause paralysis in mice, does not grow in the spinal cord, and does not cause motor neuron loss following IM injection. A panel of chimeric viruses containing sequences from IL52 and CA4231 was used to demonstrate that VP1 is the main determinant of EV-D68 neurovirulence following IM injection of neonatal SW mice. VP1 contains four amino acid differences between IL52 and CA4231. Mutations resulting in substituting these four amino acids (CA4231 residues into the IL52 polyprotein) completely abolished neurovirulence. Conversely, mutations resulting in substituting VP1 IL52 amino acid residues into the CA4231 polyprotein created a virus that induced paralysis to the same degree as IL52. Neurovirulence following infection of neonatal SW mice with parental and chimeric viruses was associated with viral growth in the spinal cord.

IMPORTANCE

Emerging viruses allow us to investigate mutations leading to increased disease severity. Enterovirus D68 (EV-D68), once the cause of rare cases of respiratory illness, recently acquired the ability to cause severe respiratory and neurologic disease. Chimeric viruses were used to demonstrate that viral structural protein VP1 determines growth in the spinal cord, motor neuron loss, and paralysis following intramuscular (IM) injection of neonatal Swiss Webster (SW) mice with EV-D68. These results have relevance for predicting the clinical outcome of future EV-D68 epidemics as well as targeting retrograde transport as a potential strategy for treating virus-induced neurologic disease.

INTRODUCTION

Enterovirus D68 (EV-D68) is a positive-sense, single-stranded RNA virus belonging to the genus Enterovirus of the Picornaviridae family. Unlike the majority of enteroviruses which spread via fecal-oral transmission, EV-D68 is similar to rhinoviruses, replicating most efficiently in the cooler temperatures of the upper respiratory tract, and causing respiratory symptoms including, sneezing, cough, runny nose and, in some cases, wheezing and difficulty breathing (1–4). EV-D68 was first isolated from children with pneumonia and bronchiolitis in California in 1962 (1) but remained a very rare cause of documented infection, with only 26 confirmed cases of EV-D68 in the United States (US) from 1970 to 2005 (5). Between 2005 and 2014, sporadic clusters of acute respiratory EV-D68 infections began to occur worldwide (6–9) and in 2014, EV-D68 caused an unprecedented outbreak in the US, with at least 1,395 confirmed cases of respiratory infection (10). A dramatic upsurge in the number of cases of acute flaccid myelitis (AFM), a disease with a striking resemblance to paralytic polio (11–14), occurred concurrently with the 2014 EV-D68 respiratory disease outbreak (15). AFM cases occurred primarily in children, many of whom had fever and upper respiratory illness prior to the onset of limb weakness, and approximately 50% of whom had EV-D68 nucleic acid in respiratory secretions (12, 16, 17). Similar large outbreaks of EV-D68 respiratory infection that were associated with AFM or neurologic complications also occurred in 2014 in Canada, Argentina, Norway, Netherlands, France, Germany, and Taiwan (18–25) and 2016 and 2018 in the US (26, 27) and elsewhere (22, 23, 28–32). The number of respiratory EV-D68 infections and AFM was much lower in 2020, presumably due to COVID-related mask usage and increased social distancing (10). Surprisingly, in 2022 EV-D68 respiratory infection recurred at levels surpassing those seen in 2014, but without an increase in AFM cases, suggesting that the respiratory and neurologic disease phenotypes are governed by distinct viral factors. EV-D68 antigen and nucleic acid have now been identified in the spinal cord of a patient who died with AFM (33). This observation, alongside epidemiological studies and detection of EV-D68 antibodies in the cerebral spinal fluid of AFM patients, has led to the recognition of EV-D68 as the main cause of the biennial spikes in AFM cases seen in the US in 2014, 2016, and 2018.

Following the association of EV-D68 respiratory disease and AFM, several animal models were developed which demonstrated that most, but not all, EV-D68 clinical isolates from the 2014 outbreak caused paralysis in neonatal and/or immunocompromised mice (34–39). Paralysis also occurs following intracerebral, intraperitoneal (IP), intranasal, and intramuscular (IM) injection of mice with EV-D68 (34). However, in our hands, IM injection provides the most consistent model of EVD68 AFM and has been used to investigate the neuropathogenesis of AFM and to evaluate several antiviral and antibody-based treatments (34, 39–42). Following IM injection of neonatal Swiss Webster (SW) mice with EV-D68 US/IL/14-18952 (IL52), a clinical isolate from the 2014 epidemic, paralysis is induced in 100% of infected mice and is associated with increased viral titer in the spinal cord as well as infection, and the subsequent loss, of motor neurons in the anterior horn (34, 40–42). Paralysis starts in the injected limb and progresses to the remaining limbs of the animal (42), consistent with initial retrograde axonal transport of the virus in the sciatic nerve followed by retrograde transsynaptic transport within the spinal cord. Progressive paralysis is also seen in human AFM, typically starting in the upper limbs, and presumably reflects axonal retrograde transport of the virus from the respiratory tract (43). Retrograde transport, involving the passage of the virus to the central nervous system (CNS) via peripheral nerves, is a common mechanism by which viruses, including other enteroviruses (poliovirus and EV-A71), enter the CNS (44–48).

The EV genome encodes a polyprotein with a single open reading frame between an untranslated region (UTR) at both 5′ and 3′ ends. Secondary structures form an internal ribosome entry site in the 5’UTR and mediate virus translation (49). The EV-D68 polypeptide contains a P1 region that encodes the structural proteins (VP1-4) and P2 and P3 regions encoding non-structural proteins (2A-C, 3A-D) that are important for virus replication. Investigations of poliovirus virulence have shown that minor sequence variations in restricted areas of the genome (in particular the 5’UTR and VP1 gene) are sufficient to account for large differences in neurovirulence between strains (50). In adult mice deficient in interferon (IFN) signaling VP1 and VP3 proteins are determinants of neurovirulence following IP infection and hematogenous spread of EV-D68 (39). In order to identify genomic regions important for neurovirulence following IM EV-D68 injection we wanted to identify molecular determinants of neurovirulence in a model more closely resembling human disease. We constructed chimeras and mutant viruses based on the 2014 isolates IL52, which causes paralysis in mice, and CA/14-4231 (CA4231), which does not, and examined their capacity to induce paralysis in immunocompetent neonatal mice. We showed that VP1 is the main determinant of viral replication in the spinal cord, motor neuron loss, and paralysis following IM infection of EV-D68.

MATERIALS AND METHODS

Cells and viruses

Rhabdomyosarcoma (RD) cells (American Type Culture Collection/ATCC, Manassas, VA) were maintained in Dulbecco’s modified Eagle’s medium (DMEM, Sigma-Aldrich, St Louis, MO) supplemented with 10% fetal bovine serum (FBS, Sigma-Aldrich) and 1X penicillin/streptomycin (Gibco, Gaithersburg, MD) at 37°C with 5% CO2. Viruses used in the described experiments were generated from infectious clones. An infectious clone of IL52 (BEI Catalog Number NR-49131, GenBank KM851230) was obtained from Raul Andino Ph.D. (University of California, San Francisco, courtesy of Ming Te Yeh Ph.D.). Upon sequencing, it was found that this clone has two amino acid changes from the Genbank IL52 polypeptide sequence (D554N and E835G). These changes were also identified by Yeh et al (38). Clinical viral isolate CA4231 (Genbank KU844181.1) was obtained from the California Department of Public Health (courtesy of Shigeo Yagi, Ph.D.). To generate an infectious clone from CA4231, viral RNA was isolated using Qiagen QIAamp Viral RNA Minikit (Qiagen, Germantown, MD) following the manufacturer’s protocol. cDNA was generated from isolated RNA using Invitrogen Superscript III (Thermo Fischer Scientific, Waltham, MA) and was amplified with EV-D68 specific primer pairs using Platinum Super-fi polymerase (Thermo Fischer Scientific), which produced blunt-ended fragments for insertion into TOPO XL-2 vectors. One Shot OmniMAX 2 T1R chemically competent cells (Thermo Fischer Scientific) were used for the transformation of the ligated vector and were subsequently streaked on plates containing ampicillin and kanamycin (Sigma-Aldrich). Resistant colonies were grown overnight in LB broth (Thermo Fischer Scientific). Plasmids were collected from resistant colonies using Qiagen QIAprep Spin Miniprep Kit and were sequenced (Quintara Biosciences, San Fransisco, CA) to identify plasmids containing correctly inserted CA4231 sequences. 4231(4241)IL52 was generated by digestion of IL52 and CA4231 infectious clones with EcoR1 and Stu 1 (cuts at EV-D68 nucleotide 4241). Ligation, transformation, and sequence verification were as described above. Infectious clones of IL52(3220)4231 and 4231(3220)IL52 were generated from IL52 and CA4231 infectious clones using a NEBuilder HiFi Assembly Cloning Kit. Ligation, transformation, and sequence verification were as described above. Point mutations were introduced into IL52 and CA4231 infectious clones by PCR using non-overlapping primers (Q5 Site Directed mutagenesis kit, New England Biolabs, Ipswich, MA). The resulting PCR product was column purified (DNA Clean and concentrator-5, ZYMO Research, Irvine, CA), and 5 µL was used in a multi-enzyme KLD reaction (New England Biolabs), which kinases, ligates, and destroys remaining plasmid template for rapid circularization. Two microliters of this reaction were transformed into DH5α highly competent cells (Thermo Fisher Scientific). Plasmids were isolated and screened by sequencing as described above. Untranslated regions from IL52 and CA4231 were inserted into infectious clones using conserved cut sites Pml1 (40 bp) and Nhe1 (818 bp). To generate viruses from infectious clones, plasmids were linearized using Sal 1. RNA was then synthesized (HiScribe synthesis and Monarch cleanup kits, New England Biolabs), quantified by NANOdrop, and transfected into RD cells (Lipofectamine 3000, Invitrogen). Infected cells were grown/expanded in RD cells at 33°C and 5% CO₂ until most cells were dead or dying. Cell debris was removed by ultracentrifugation, and the titer of the resulting stocks was determined by TCID50. All infectious clones and subsequent progeny viruses were sequenced and found to be identical to the original viruses.

Murine model

Pregnant Swiss Webster females were purchased from Envigo (Indianapolis, IL). Interferon deficient [mitochondrial antiviral signaling protein (MAVS −/−)] mice bred on a pure C57BL/6 background were obtained as a generous gift from Michael Gale Ph.D. (University of Washington) as homozygous breeding pairs whose genotypes were verified using previously described methods. Male and female mouse pups were injected on days 1–2 of life (weight >1.5 g but <2 g) by IM injection in the left rear limb with 1000 TCID50 virus (10 µL volume). Litter sizes were maintained at 8–12 pups per mother. When possible, multiple litters were combined and pups were distributed between mothers. Mice were scored for paralysis using a previously validated scoring system (34, 39–41) where each limb is scored 0–3 (0 = normal function, 3 = complete paralysis). Alternatively, muscle and spinal cord were harvested at 6 days post-infection (DPI) for determination of viral load by TCID50 assay. For paralysis curves and day 10 paralysis scores, where mice died or were sacrificed, the highest paralysis score reached by each mouse was recorded and carried forward until day 10. The paralysis scoring system has previously been described (41). Scorers were blinded to which viruses were used in each group.

TCID50 assay

RD cells were plated at 4,000 cells/well in 180 mL of media in a 96-well flat-bottomed plate and were incubated overnight at 33°C/5% CO₂. Spinal cord and muscle tissue were collected in 300 µL of PBS and homogenized using a BeadBug tissue homogenizer. Serial 10-fold dilutions of tissue lysate were prepared, and 20 µL of each dilution was added to the cells. After 2 h, the wells were rinsed and media replenished (200 mL per well). The plate was placed in an incubator for a week, and the wells were assessed daily for signs of CPE. TCID50 values were calculated using the Reed-Muench method (34, 39–41).

Immunohistochemistry

Eight days following the infection, animals were sacrificed and spinal cords were dissected from the spinal column in ice-cold PBS. Spinal cords were fixed in 4% paraformaldehyde for 24 h and cryoprotected in 30% sucrose for 24–48 h. The lumbar enlargement was dissected from the spinal cord and then frozen in OCT. Blocks were transversely sectioned at 20-micron thickness using a Leica CM3050S cryotome prior to immunohistochemical staining. Antigen retrieval was performed and sections were stained with Anti-NeuN Rat mAb ab279297 (Abcam: ab279297 used at 1:100). A Leica confocal microscope with 10× objective was used to image spinal cord sections. The number of motor neurons in the right and left anterior horns of sections from 2 to 3 animals were counted.

Statistical analysis

GraphPad Instat (unpaired t-tests) and PRISM [analysis of variance (ANOVA)] software were used for statistical analyses.

RESULTS

IL52, but not CA4231, induces paralysis following IM injection of mice deficient in interferon signaling

IL52, but not CA4231, induces paralysis following IC injection of neonatal SW mice (34). However, CA4231 has similar neurovirulence to IL52 following IP injection of immunocompromised Type 1 IFN receptor knock-out mice (39). This suggests that differences in neurovirulence between IL52 and CA4231 may be determined by differences in IFN responses or differences in neuroinvasion (hematogenous or neural pathways of entry to the CNS). We used MAVS knock-out (ko) mice to demonstrate that following IM injection, IL52, but not CA4231 causes paralysis in mice deficient in IFN signaling (Fig. 1) suggesting that differences in neurovirulence between IL52 and CA4231 following IM injection lie in their different abilities to enter the CNS by retrograde transport from the muscle.

Fig 1.

Following IM injection IL52, but not CA4231, induces paralysis in MAVS ko mice. Two-day-old MAVS ko mice were infected with 1000 TCID50 EV-D68 strains IL52 (N = 3) and CA4231 (N = 4) by IM inoculation. The graph shows the average paralysis score of infected mice at DPI 4 and 8. The paralysis score of individual mice was based on a previously established scoring system where each limb is given a score of 0 (normal function)–3 (complete paralysis). Error bars are SEM. Statistical analysis was performed using PRISM (unpaired t-test).

IL52, but not CA4231, induces paralysis following IM injection of mice that is associated with viral growth in the spinal cord and loss of motor neurons

Consistent with our previous studies (34, 40, 50), following IM injection of 1000 TCID50 IL52 100% of animals become paralyzed (Fig. 2A and B), which peak scores ~6 at 9 DPI. Seventy percent of infected mice develop complete paralysis in both hind limbs (total paralysis score 6), and 15% of infected mice develop complete hind limb paralysis which also spreads to the upper limbs (Fig. 2B). The remaining 15% of infected animals have complete paralysis in the injected limb but retain some movement in the contralateral (non-injected) hind limb. Paralysis following IM injection of IL52 is associated with viral growth in both the spinal cord and injected muscle (Fig. 2C) and loss of motor neurons from the anterior horns of the spinal cord (Fig. 2D and E). In contrast, CA4231 does not cause paralysis following IM injection (Fig. 2A and B), no virus is detected in the spinal cord (Fig. 1C) and there is no loss of motor neurons compared to mock-infected controls. CA4231 does grow in the muscle following IM injection (Fig. 2C) to a similar degree as IL52 demonstrating equivalent replication fitness.

Fig 2.

IL52 induces paralysis in mice that is associated with viral growth in the spinal cord and loss of motor neurons. Two-day-old SW mice were infected with 1000 TCID50 EV-D68 strains IL52 (N = 17) and CA4231 (N = 15) by IM inoculation. (A) The graph shows the average paralysis over time of mice infected with IL52 (black circles) and CA4231 (gray circles). The paralysis score of individual mice was based on a previously established scoring system where each limb is given a score of 0 (normal function)–3 (complete paralysis). Error bars are SEM. (B) The scatter plot shows the paralysis scores for individual IL52 (black circles, N = 17) and CA4231 (gray circles, N = 15)-infected mice at DPI 10. The mean paralysis score is shown. Error bars are SEM. Statistical analysis was performed using PRISM (unpaired t-test). (C) In separate experiments, the injected muscle and spinal cords were harvested from IL52 (N = 7) and CA4231 (N = 6)-infected mice at DPI 6. The scatter plot shows titers for individual mice as well as the average viral load per gram of tissue for each group. Error bars are SEM. Statistical analysis was performed using PRISM (unpaired t-test), ns = not significant. (D and E) Spinal cords were harvested from mock-, CA4231-, and IL52-infected mice (N = 6). Immunohistochemical analysis of spinal cord sections demonstrated that the spinal cord motor neurons (large staining cells in white box) are present in mock- and CA4231-infected animals but are completely absent from the anterior horns of mice infected with IL52. The images (D) are representative sections. Graph (E) shows the motor neuron counts for individual mice as well as the average (for each group) number of motor neurons in the right (R) and left (L) lumbar enlargements (N = 6). Error bars are SEM. Statistical analysis was performed comparing the total number of motor neurons (both enlargements) using PRISM (unpaired t-test).

The non-structural proteins and 3’UTR do not account for differences in neurovirulence between IL52 and CA4231

IL52 and CA4231 share 98% sequence homology. There are 13 nucleotide differences in the 5’UTR, 1 nucleotide difference in the 3’UTR (Fig. 3A), and 12 amino acid differences in the polyprotein (Fig. 3B) between the two viruses. The amino acid sequences of the structural proteins VP4 and VP2, and the non-structural proteins 2C, 3B, and 3C are identical between both viruses (Fig. 3B). In order to determine which changes are associated with neuropathogenesis we used reverse genetics to create chimeras between IL52 and CA4231 and to introduce point mutations into each virus. The virus 4231(4241)IL52 contains CA4231 nucleic acid sequences for the 5’UTR through 2B (5’UTR, VP1-4, 2A, 2B) and IL52 sequences for 2C through the 3’UTR (2C, 3A-D, 3’UTR). This virus produced no paralysis in mice following IM injection (Fig. 3C) demonstrating that sequences upstream of 2C determine neurovirulence. The chimera 4231(3220)IL52 contains CA4231 sequences for the 5’UTR and VP1-4 and IL52 sequences for non-structural protein 2A through the 3’UTR (2 A-2C, 3A-3D, and 3”UTR). 4231(3220)IL52 also does not produce paralysis in mice following IM injection, further localizing the determinant of neurovirulence to the structural proteins and/or the 5’UTR. Supporting these findings, the viruses 2A/IL52, 2B/IL52, and 2AB/IL52 that contain a CA4231 residue in 2A (K886R), 2B (D1038N), or both proteins (K886R/D1038N) in an IL52 background, all have similar paralysis scores to IL52 at 10 DPI (Fig. 3C). Similarly virus 3A/IL52, which contains a CA4231 residue at position in the 3A protein (S1496N), in an IL52 background, induces similar paralysis to IL52 (Fig. 3C and D). Taken together these results indicate that the non-structural proteins and 3’UTR do not account for differences in neurovirulence between IL52 and CA4231.

Fig 3.

The non-structural proteins do not determine neurovirulence following infection of mice with IL52. This figure contains data from parental viruses (Fig. 2) that is presented for comparison. (A) IL52 and CA4231 have 13 nucleotide differences in the 5’UTR and 1 nucleotide difference in the 3’UTR. (B) The IL52 and CA4231 polyproteins have 12 amino acid differences: 2 in VP3; 4 in VP1; 1 in 2A, 2B, and 3A; and 3 in 3D. (C) Diagram of parental strains (IL52 and CA4231), engineered chimeras [4231(4241)IL52 and 4231(3220)IL52], and viruses with individual CA4231 amino acid residues (2A, 2B, 2A and 2B, and 3A) in an IL52 background. Individual amino acid differences between IL52 and CA4231 are shown by black arrows. Proteins and UTR sequences from IL52 are shown in black. Proteins and UTR sequences from CA4231 are shown in gray. Individual mutations where CA4231 amino acid residues replace IL52 residues are marked with a gray line in an otherwise black segment. The white regions have no amino acid differences between IL52 and CA4231 (VP4, VP2, 2C, 3B, and 3C). Two-day-old SW mice were infected by IM injection with 1000 TCID50 EV-D68 strains IL52 (N = 17), CA4231 (N = 15), 4231(4241)IL52 (N = 16), 4231(3220)IL52 (N = 24), 2A/IL52 (N = 8), 2B/IL52 (N = 8), 2A2B/IL52 (N = 11), and 3A/IL52 (N = 8). The average paralysis score (DPI 10) and SEM are shown to the right of the genome diagrams. (D) The scatter plot shows the individual paralysis scores of each mouse as well as the average paralysis for each group. IL52-infected mice are represented by black circles, CA4231-infected mice are represented by gray circles, and chimera/mutant-infected mice are represented by black diamonds. Error bars are SEM. Statistical analysis was performed using PRISM (unpaired t-test). Paralysis scores for 2A/IL52, 2B/IL52, 2A2B/IL52, and 3A/IL52 were not significantly different for those of IL52.

VP1 is the main determinant of neurovirulence between IL52 and CA4231

Having ruled out EVD68 non-structural proteins and 3’UTR as significant determinants of the paralytic phenotype this left the 5’UTR, VP1, or VP3 as potential candidates (VP2 and VP4 have identical amino acid sequences between IL52 and CA4231). Mutant viruses were constructed (Fig. 4A) that contained (i) 4 CA4231 VP1 amino acid residues (L553I, N554D, T650A, E835K) in an IL52 background (VP1/IL52), (ii) 2 CA4231 VP3 amino acid residues (R358H, V470I) in an IL52 background (VP3/IL52), or (iii) the 5’UTR from CA4231 in an IL52 background (UTR/IL52). At 10 DPI VP1/IL52 infected mice had no paralysis—identical to CA4231 (Fig. 4B)—demonstrating that VP1 is a major determinant of paralysis between IL52 and CA4231. In support of these findings both VP3/IL52 had similar paralysis scores to IL52 at DPI 10 and UTR/IL52 infected mice had somewhat increased paralysis scores to IL52 at 10 DPI (Fig. 4B). Paralysis following IM injection of IL-52, UTR/IL52, and VP3/IL52 is associated with viral growth in both the spinal cord and injected muscle (Fig. 4C). In contrast, VP1/IL52 (as well as CA4231) infected mice do not have detectable virus in the spinal cord (Fig. 4C). All viruses grow in the muscle following IM injection (Fig. 4D) to a similar degree as IL52, demonstrating equivalent replication fitness. VP3/IL52 muscle titer was significantly increased compared to IL52 but it is unclear what biological relevance this increased titer might have.

Fig 4.

VP1 is the main determinant of neurovirulence following infection of mice with EV-D68. This figure contains data from parental viruses (Fig. 1) that is presented for comparison. (A) Diagram of engineered and parental (same data as shown in Fig. 1 but included as a comparison) viruses. Engineered viruses include viruses with individual CA4231 amino acid residues in VP1 (VP1/IL52 contains 4 CA4231 amino acid residues) and VP3 (VP3/IL52 contains 2 CA4231 residues), or the CA4231 5’UTR (UTR/IL52) in an IL52 background. Individual amino acid differences between IL52 and CA4231 are shown by black arrows. Proteins and UTR sequences from IL52 are shown in black. Proteins and UTR sequences from CA4231 are shown in gray. Individual mutations where CA4231 amino acid residues replace IL52 residues are marked with a gray line in an otherwise black segment. The white regions have no amino acid differences between IL52 and CA4231 (VP4, VP2, 2C, 3B, and 3C). Two-day-old SW mice were infected by IM injection with 1000 TCID50 EV-D68 strains IL52 (N = 17), CA4231 (N = 15), VP1/IL52 (N = 16), VP3/IL52 (N = 16), and UTR/IL5 (N = 24). The average paralysis score at DPI 10 and SEM are shown to the right of the genome diagrams. (B) The scatter plot shows the paralysis score of individual mice at DPI 10 as well as the average paralysis score. IL52-infected mice are represented by black circles, CA4231-infected mice are represented by gray circles, and chimera/mutant-infected mice are represented by black diamonds. Error bars are SEM. Statistical analysis was performed using InStat (ANOVA). Paralysis scores for VP3/IL52 and UTR/IL52 were not significantly different from those of IL52. (C and D) In a separate experiment spinal cord and the injected muscle from infected mice were harvested at DPI 6. The scatter plots show viral loads for individual mice as well as the average virus loads per gram of tissue in the spinal cord (C) and muscle (D) for IL52, CA4231, VP1/IL52, VP3/IL52, and UTR/IL5 (N = 8 for all viruses). IL52-infected mice are represented by black circles, CA4231-infected mice are represented by gray circles, and chimera/mutant-infected mice are represented by black diamonds. Error bars are SEM. Statistical analysis was performed using InStat (ANOVA). Spinal cord titers for VP3/IL52 and UTR/IL52 were not significantly different for those of IL52. Muscle titers for VP1/IL52 and UTR/IL52 were not significantly different for those of IL52.

IL52 sequences confer neurovirulence to CA4231

Having shown that the substitution of 4 amino acids from CA4231 VP1 into IL52 can block paralysis we investigated whether the substitution of IL52 sequences into CA4231 could confer the paralytic phenotype. Mutant viruses were constructed that contained (i) the 5’UTR and VP1-4 of IL52 with 2A-3C and the 3’UTR from CA4231 [IL52(3220)4231], (ii) the 5’UTR, 2 IL52 VP3 amino acid residues (H358R, I470V) and 4 IL52 VP1 amino acid residues (I553L, D554N, A650T, K835E) in an CA4231 background (UTRVP3VP1/4231), (iii) 4 IL52 VP1 amino acid residues (I553L, D554N, A650T, K835E) in an CA4231 background (VP1/4231) (Fig. 5A). Mice infected with 1000 TCID50 of IL52(3220)4231, UTRVP3VP1/4231, and VP1/4231 all developed paralysis (Fig. 5B). Paralysis induced by UTRVP3VP1/4231, and VP1/4231 was similar to that induced by IL52 indicating that the Il52 VP1 is sufficient for neuropathogenesis. Interestingly, a significantly higher average paralysis score (8.5) was seen in mice infected with IL52(3220)4231 compared to IL52 (5.9). Analysis of daily paralysis scores revealed that increased paralysis at DPI 10 followed an earlier onset of paralysis: at 3 DPI the average paralysis score of mice infected with IL52(3220)4231 was 2.5, compared to 0.7 for IL52 (Fig. 6).

Fig 5.

IL52 sequences confer neurovirulence to CA4231. This figure contains data from parental viruses (Fig. 1) for comparison. (A) Diagram of parental viruses and engineered chimeras and viruses with specific IL52 nucleotide and amino acid residues in a CA4231 background. Individual amino acid differences between IL52 and CA4231 are shown by black arrows. Proteins and UTR sequences from IL52 are shown in black. Proteins and UTR sequences from CA4231 are shown in gray. Individual mutations where CA4231 amino acid residues replace IL52 residues are marked with a black line in an otherwise gray segment. The white regions have no amino acid differences between IL52 and CA4231 (VP4, VP2, 2C, 3B, and 3C). Two-day-old SW mice were infected by IM injection with 1000 TCID50 EV-D68 strains IL52 (N = 17), CA4231 (N = 15), IL52(3220)4231 (N = 11), UTRVP3VP1/4231 (N = 24), and VP1/4231 (N = 17). The average paralysis score at DPI 10 and SEM are shown to the right of the genome diagrams. (B) The scatter plot shows the paralysis scores for each mouse as well as the average paralysis score for each group at DPI 10. IL52-infected mice are represented by black circles, CA4231-infected mice are represented by gray circles, and chimera/mutant-infected mice are represented by black diamonds. Error bars are SEM. Statistical analysis was performed using InStat (ANOVA). Paralysis values for UTRVP3VP1/4231 and VP1/IL52 were not significantly different from those of IL52.

Fig 6.

Increased neurovirulence in IL52(3220)4231 reflects higher paralysis scores at DPI 3 and higher spinal cord titers at DPI 6. This figure contains data from parental viruses (Fig. 1) as a comparison. Two-day-old SW mice were infected with 1000 TCID50 EV-D68 strains IL52 (N = 17) and IL52(3220)4231 (N = 11) by IM inoculation. (A) The graph shows the average paralysis of mice infected with IL52 (black circles) and IL52(3220)4231 (gray circles). The paralysis score of individual mice was based on a previously established scoring system where each limb is given a score of 0 (normal function)–3 (complete paralysis). Error bars are SEM. (B) Scatter plot showing individual paralysis scores for IL53 (black circles) and IL52(3220)4231 (black diamonds) as well as average scores for the group at DPI 3. Error bars are SEM. Statistical analysis was performed using PRISM (unpaired t-test). (C) Scatter plot showing individual viral spinal cord titers for IL52 (black circles) and IL52(3220)4231 (black diamonds), as well as the average titer for each group. Error bars are SEM. Statistical analysis was performed using PRISM (unpaired t-test).

DISCUSSION

Historically, EV-D68 was the causative agent of rare cases of human respiratory disease. Recently, EV-D68 has been associated with more severe respiratory illness as well as the paralytic disease AFM, which became emergent in 2014. Many, but not all, EV-D68 strains isolated from patients in 2014 produced paralytic disease in mouse models of AFM (34–39). To determine if increased neurovirulence is the result of genetic differences between virus isolates we compared two isolates from the 2014 epidemic that are genetically similar but which differed in their ability to cause paralysis following IM injection of neonatal SW mice. IL52 causes paralysis in mice following IM injection of the left hind limb. Mice develop paralysis in the injected limb beginning around 3–4 DPI with IL-52 and paralysis spreads to the contralateral limb in the majority of cases. Forelimb paralysis can also occur but is less common. In contrast to IL52, CA4231 does not cause paralysis in this model of AFM. IL-52 and CA4231 differ by 13 nucleotides in the 5’UTR, 1 nucleotide in the 3’UTR, and 12 amino acids in the viral polyprotein. Chimeras 4231(4241)IL52 and 4231(3220)IL52 contain 5’ sequences derived from CA4231 and 3’ sequences from IL-52. Specifically, 4231(4241)IL52 contains CA4231 sequences up to bp 4241 (the 5’UTR and viral proteins VP1-4, 2A, and 2B) and IL52 sequences downstream of bp 4241 (viral proteins 2C, 3A-3D, and the 3’UTR), whereas 4231(3220)IL52 contains the CA4231 sequences through bp 3220 (5’UTR and viral proteins VP1-4) and IL52 sequences downstream of bp 3220 (viral proteins 2A-C, 3A-D, and the 3’UTR). Neither 4231(4241)IL52 nor 4231(3220)IL52 cause paralysis in mice, demonstrating that the 5’UTR or structural proteins (VP1-4) of EV-D68 are vital for neuropathogenesis following IM injection. In support of this, IL52 containing CA4231 amino acid residues in the 2A, 2B, or 3A regions were similar in neuropathogenesis to IL52. IL52 containing the 5’UTR or VP3 from CA4231, also caused similar amounts of paralysis to IL52, ruling out the 5’UTR and VP3 as major determinants of neuropathogenesis. VP2 and VP4 have identical amino acid sequences between IL52 and CA4231. Taken together these results suggest that the four amino acid changes in VP1 between IL52 and CA4231 determine differences in neurovirulence following IM injection of mice with these viruses. Insertion of IL52 VP-1 amino acids in a CA4231 background restored the paralytic phenotype, confirming the importance of VP1 in neuropathogenesis.

These results differ from those obtained by Yeh et al. (38) whose work indicated that a single amino acid substitution in VP3 was the main determinant of EV-D68 paralysis. Yeh et al. did identify VP1 as a minor determinant of neuropathogenesis in their model (38) and VP1 has been implicated in neuropathogenesis following murine EV-A71 infections (51). Enterovirus VP1 forms the vertices of the viral capsid, is the most external and immuno-dominant of the picornavirus capsid proteins (52), and together with VP3, interacts with host cell sialic acids to facilitate virion attachment to some host cells (53).

It should be noted that the model used by Yeh et al. involved immuno-deficient Tg21/IFNR-ko mice (54) and IP delivery of virus (38) and that CA4231 is neuropathogenic in this model. This suggests that the CA4231 vs IL52 differences in neuropathogenesis following IM injection are specific to the retrograde transport of the virus to the spinal cord. It has been previously suggested that in humans EV-D68 spreads from the lungs into the bloodstream to establish viremia and then enters the CNS by direct hematogenous seeding or following replication in muscle fibers, translocation across the neuromuscular junction and retrograde transport within the motor axon (55). However, given the increased prevalence of upper limb paralysis and instances of diaphragm paralysis following respiratory infection in humans, and progressive paralysis from the injected limb in mice (42), we hypothesize that muscle infection followed by retrograde transport is the predominant route of CNS infection in humans. Direct hematogenous seeding would be expected to result in a more even distribution of paralysis. This makes a retrograde transport model of AFM an attractive tool for determining factors that affect viral transport to the spinal cord. However, host mechanisms to limit viral growth at the site of infection and hematogenous spread are also likely to impact disease.

Hixon et al. previously demonstrated that virus growth in the spinal cord was associated with paralysis following IC, IM, and IN infection of neonatal SW mice with EV-D68 isolate US/MO/14-18947 (MO-47) and following IM infection neonatal SW mice with IL-52 (34, 39, 42). Virus was detected within spinal cord motor neurons at 2–3 DPI (prior to paralysis) and corresponded to subsequent motor neuron loss (34, 42). Results presented here are consistent with those studies and further demonstrate the association of viral growth in the spinal cord with paralysis and loss of motor neurons. Neither CA4231 nor any of the none-paralytic chimeras had detectable virus in the spinal cord at 6 DPI when viral load is expected to be highest (34), although all replicated to the same extent in infected muscle demonstrating similar overall viral fitness. In contrast, all viruses that induced paralysis had detectable virus in the spinal cord.

Insertion of 4 IL-52 VP1 amino acids into a CA4231 background (VP1/4231) or insertion of the IL-52 5’UTR, as well as 4 IL52 VP1 amino acids and 2 IL-52 VP3 amino acids into a CA4231 background, was sufficient to change a non-paralytic virus (CA4231) to one with the same paralytic phenotype as IL52. However, a virus containing the whole front of IL-52 (5’NTR and VP1-4) and the back half of CA4231 (IL52-3220-4231) was significantly better at causing paralysis than IL52 (paralysis score 8.5 v 5.9). This suggests there may be additional silent mutations between IL52 and CA4231 (VP1-4) that are important for pathogenesis even though they do not result in an amino acid change. Silent mutations can affect protein levels or conformation by modifying mRNA stability, miRNA binding sites, translation efficiency, or splicing regulatory sites (56, 57), and several studies have recently demonstrated that silent mutations may potentially be harmful (58, 59).

Identification of VP1 as the major determinant of neuropathogenesis following IM injection will help inform the rational design of EV-D68 vaccines, antivirals, and monoclonal antibody-based therapies.

ETHICAL APPROVAL

All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Colorado (protocol # 00075). Mice were humanely sacrificed in cases of severe disease (typical paralysis scores 10–12) in which forelimb paralysis prevented proper feeding and resulted in significant weight loss (20%).

DATA AVAILABILITY

The authors confirm that the data supporting the findings of this study are available within the article. Viruses and plasmids are available upon request.