Author Correction: Pharmacological perturbation of CXCL1 signaling alleviates neuropathogenesis in a model of HEVA71 infection

Hand, foot and mouth disease (HFMD) caused by Human Enterovirus A71 (HEVA71) infection is typically a benign infection. However, in minority of cases, children can develop severe neuropathology that culminate in fatality. Approximately 36.9% of HEVA71-related hospitalizations develop neurological complications, of which 10.5% are fatal. Yet, the mechanism by which HEVA71 induces these neurological deficits remain unclear. Here, we show that HEVA71-infected astrocytes release CXCL1 which supports viral replication in neurons by activating the CXCR2 receptor-associated ERK1/2 signaling pathway. Elevated CXCL1 levels correlates with disease severity in a HEVA71-infected mice model. In humans infected with HEVA71, high CXCL1 levels are only present in patients presenting neurological complications. CXCL1 release is specifically triggered by VP4 synthesis in HEVA71-infected astrocytes, which then acts via its receptor CXCR2 to enhance viral replication in neurons. Perturbing CXCL1 signaling or VP4 myristylation strongly attenuates viral replication. Treatment with AZD5069, a CXCL1-specific competitor, improves survival and lessens disease severity in infected animals. Collectively, these results highlight the CXCL1-CXCR2 signaling pathway as a potential target against HFMD neuropathogenesis. Neurological complications are frequently observed in severe Hand, Foot and Mouth Disease but the underlying causes are unknown. Here, the authors report infected astrocytes release CXCL1 which activates neuronal CXCR2 to cause neuropathogenesis.

H uman Enterovirus A71 (HEVA71) is a small, positivesense, single-stranded RNA virus that was first isolated in 1969 from patients with severe central nervous system diseases 1 . It has since emerged as one of the most important causative agents of hand, foot and mouth disease (HFMD), a serious global public health threat with epidemic outbreaks in the Asia-Pacific regions with high annual disability-adjusted life years (DALYs) 2,3 . Despite its severity, there is a paucity of treatments currently available to affected patients 4 . While usually a benign and self-limiting infection, a substantial portion of HEVA71-infected children under 3 years of age continue to develop serious neurological deficits including aseptic meningitis, brain stem encephalitis and poliomyelitis-like acute flaccid paralysis that can culminate in permanent neurological sequelae or even fatality 5,6 . Post-mortem studies confirmed the presence of HEVA71 infection in the central nervous system (CNS), particularly in the medulla oblongata and spinal cord, indicating that the neurological complications are linked to viral replication in the CNS 7,8 . Dysregulated cytokine production is strongly linked to severe HEVA71 infections with neurological complications 9 . In particular, cytokines modulated by specific HEVA71 proteins are known to be hijacked by the virus to subvert host responses 10 . For example, the 2A protease reduces type I IFN receptor (IFNAR1) expression, preventing interferon (IFN)-facilitated antiviral responses in skeletal muscle cells 11 . Similarly, the 3C protease suppresses mitogenactivated protein kinase (MAPK) signaling to reduce the expression of anti-viral pro-inflammatory cytokines in intestinal epithelial cells 12 . However, the identities of CNS-specific cytokines and the viral proteins directly responsible for modulating their production remains undefined 13,14 . There remains a critical knowledge gap surrounding the relationship between these cytokines and their contributions to neurological complications in severe cases and fatality. Insights derived from studies using non-neuronal cells have proven ineffective against neurological symptoms and the availability of pertinent cell-based models that efficiently recapitulate the HEVA71 replication cycle in neurons is sorely needed to clarify this relationship.
Here, we demonstrate that HEVA71 infection of primary rat neurons and astrocytes recapitulates neuronal and astrocytic damage observed in the CNS of infected animals [15][16][17] and humans [18][19][20] . We show that viral protein VP4 triggers secretion of CXCL1 from infected astrocytes, which then acts via its receptor CXCR2 to enhance HEVA71 replication in neurons. Elevated levels of CXCL1 in cerebrospinal fluids (CSF) closely correlated with disease severity in HEVA71-infected mice, and were only detected in the CSF from severe HFMD patients with neurological complications. Using complementing strategies, we demonstrated that attenuation of neuronal CXCL1 signaling strongly suppresses viral replication. Exploiting these findings, we administered AZD5069, a selective CXCR2 antagonist, into HEVA71-infected mice and demonstrated its effectiveness in significantly lowering disease severity, extending viability and reducing viral load despite persisting elevated CXCL1 levels. Taken together, our results demonstrated that blocking CXCL1 signaling during acute HEVA71 infection can be an effective therapeutic approach to manage the effects of neuropathogenicity in severe cases of HFMD.
CPE and cell death can occur independently of productive viral infection 26 . Lack of CPE after viral infection could be attributed to lower viral infectivity 26 . To confirm if HEVA71, CA6 and CA16 replicate in neuronal cultures, we examined for the presence of viral double-stranded RNA (dsRNA), an intermediate RNA species synthesized during viral replication 4 . In parallel, neurons were stained for the neuronal microtubule-associated protein (Tau) to evaluate axonal integrity 29 . In neuronal cultures exposed to UI HEVA71, few neurons showed perinuclear dsRNA staining with axon disintegration at all time points (Fig. 1a). In comparison, strong dsRNA signals were detected in HEVA71infected neurons at 6 hpi when axonal disintegration also became discernible (Fig. 1a, arrowheads). In addition, approximately 80% of HEVA71-infected neurons were positive for dsRNA at 6 hpi (Fig. 1b). These changes were also seen with HEVA71 infections at higher MOIs ( Supplementary Fig. 2a). Strikingly, neither dsRNA staining nor axon disintegration were seen in CA6 or CA16-infected neuronal cultures ( Supplementary Fig. 2b). These results indicate that, of the Enteroviruses tested here, only HEVA71 can replicate in neurons and induce CPE.
Productive infection of HEVA71 in infected cultures was further evidenced by the progressive rise in viral RNA and protein (VP0/VP2) levels and the release of infectious HEVA71 viral particles into the culture media (Fig. 1c, d; Supplementary  Fig. 3a). Notably, peak viral titers were similar to those obtained from infected human rhabdomyosarcoma (RD) cells, a common cell line used for studying HEVA71 infection, albeit obtained at a slower replication rate ( Fig. 1d; Supplementary Fig. 3b) 27 . Significant levels of apoptosis, a feature previously reported for other HEVA71 cell infection models, was also evident in HEVA71-infected neuronal cultures ( Fig. 1e; Supplementary  Fig. 3c) [30][31][32] . Interestingly, no differences in rates of replication between the S41 and C2 strains of HEVA71 were observed in neuronal cultures ( Fig. 1d; Supplementary Fig. 3a, b) despite the reported differences in clinical outcomes (Supplementary Table 1). Collectively, these results strongly support the use of primary neuronal cultures to study the mechanisms of HEVA71 neuropathogenesis.
HEVA71 infection in neuronal cultures triggers massive release of CXCL1 from astrocytes. HEVA71 infection elicits massive production of various cytokines 13,33,34 . However, their origins and contribution to infection outcomes in the CNS is poorly understood. To elucidate this, we tested culture supernatants of HEVA71-infected neuronal cultures against 23 cytokines (Supplementary Table 2). Of these, only 5 were highly upregulated. Four of the 5 cytokines (MCP-1, MIP1α, RANTES and CXCL1) are thought to originate from astrocytes and microglia 11,35 . Three of these (MCP-1, MIP1α, RANTES) peaked at time points corresponding to maximum virus production (24 hpi) (Fig. 2a). Strikingly, highest levels of CXCL1, the most abundant cytokine detected, was sustained between 24 to 48 hpi. At higher MOIs, cytokines peaked faster, again coinciding with the timing of maximum viral titers for the respective MOIs ( Supplementary Fig. 4). The last cytokine, neuronal G-CSF 36 , showed highest levels only at 48 hpi (Fig. 2a).
Neuronal cultures are known to contain some proportion of astrocytes and the finding that cytokines of astrocytic origin were released is not unexpected. Our cultures contain about 70% neurons (MAP2 + ) and 30% astrocytes (GFAP + ), similar to other reports ( Supplementary Fig. 5a) 37 . To further confirm the astrocytic origin of the cytokines, we infected highly enriched (99%; Supplementary Fig. 5b) astrocyte cultures with HEVA71. Confirming its replication in these cultures, increasing levels of dsRNA, viral protein synthesis as well as viral particles could be detected from 6 hpi onwards (Fig. 2b, c; Supplementary Fig. 5c-e). Similar to other HEVA71 cell infection models [30][31][32] , infected astrocytes undergo significant apoptosis (Supplementary Fig. 5f- Table 2). To further determine if CXCL1 could be similarly elicited during infections in vivo, mice were infected with HEVA71 as previously described 38 . Animals intracranially infected with S41 began showing increased clinical scores and mortality from 4 dpi (Fig. 3a, b). Noteworthy, hindlimb paralysis (indicative of HEVA71-induced CNS damage) contributed primarily to the high clinical scores documented from 4 dpi onwards (Supplementary Table 3) 26 . Indeed, Nissl staining revealed populations of dead neurons in both hippocampus and prefrontal cortex of S41infected mice, coincident with the presence of viral antigens ( Supplementary Fig. 6a, b; indicated by arrowheads). Furthermore, high viral titers could be measured (Fig. 3c). Meanwhile, UI-infected mice exhibited none of these adverse features. Thus, HEVA71 replication in the brains of these mice contributes to CNS damage and eventual death.
To examine if CXCL1 is released upon infection of these animals, we assayed sera and CSF collected from S41-or UIinfected mice from 0 to 7 dpi. In S41-infected mice, serum CXCL1 levels rose slightly but CSF levels of the chemokine rapidly increased from 4 dpi onwards (Fig. 3d, e). This coincided temporally with the onset and progressive worsening of mortality and clinical scores in these animals. Of note, no changes were observed in CXCL1 levels in sera or CSF of UI controls throughout the 7 day monitoring period (Fig. 3d, e).
To further clarify the source of CXCL1, we stained neuronal cultures against CXCL1 in combination with Tau (neurons) or GFAP (astrocytes). Only infected dsRNA-positive astrocytes expressed CXCL1 ( Supplementary Fig. 7, indicated by arrows). Confirming this, only astrocytes in brains of S41-infected mice expressed CXCL1 while the chemokine was noticeably absent in neurons (NeuN + ) and microglia (IBA1 + ) ( Supplementary Fig. 8a, b; indicated by arrowheads). Meanwhile, CXCL1 was not detected in all 3 cell types in UI controls.
CXCL1 is a known chemoattraction for neutrophils 39,40 . We further examined if the observed neuropathology could be caused by infiltrating neutrophils as reported in other viral infections [41][42][43] . Neonatal mice were intracranially infected with HEVA71 as before and then treated with isotype IgG (IgG) or anti-Ly6G antibodies (Ly6G-Ab) 44,45 at 1 mg/kg once daily till 5 dpi. Progressive mortality and increase in clinical scores were similar between IgG-and Ly6G-Ab-treated S41-infected mice, as were viral titers at 7 dpi (Supplementary Fig. 9a-c). Likewise, similar increase in CSF CXCL1 levels occurred from 4 dpi onwards in both groups of S41-infected mice while sera CXCL1 levels remain unchanged ( Supplementary Fig. 9d, e). As expected, UI controls remained healthy and no changes in CXCL1 levels were seen throughout the monitoring period ( Supplementary  Fig. 9a-e).
Fluorescence-activated cell sorting analyses (FACS) of dissociated cells from infected brain areas doubly labeled for CD45 and Ly6G, markers for neutrophils, showed very low neutrophil infiltration into these areas even in the absence of treatment with anti-Ly6G antibodies ( Supplementary Fig. 9f). In comparison, FACS of infected hindlimb muscles from the same animals contained elevated populations of these cells ( Supplementary  Fig. 9f). Taken together, these results demonstrate that neutrophils do not contribute to CNS damage during HEVA71 infection and reinforces the release of CXCL1 from astrocytes as an important cause of CNS damage and eventual mortality. CXCL1 release in HFMD patients has not been studied. We measured CXCL1 levels in sera and CSFs obtained from healthy individuals, patients suffering from non-HFMD-related viral infections (JEV) with CNS neurological complications 46 , as well as from HFMD patients (HEVA71, CA6 or CA16) exhibiting varying degrees of clinical severity 47,48 . All were under 65 months of age at the point of infection (Supplementary Fig. 10). Only CSFs from severe HEVA71 HFMD patients contain significantly elevated CXCL1 levels (Fig. 3f), which were similar to those detected in our infected neuronal cultures and mice. Supporting our earlier findings that CA6 and CA16 did not infect neurons, only low levels of CSF CXCL1 were present in patients severely infected by these viruses. Thus, elevated levels of CXCL1 is strongly linked to the neuropathogenesis observed in severe HEVA71 infections in humans.
CXCL1 enhances HEVA71 replication in neuronal cultures. CXCL1 binds to the neuronal CXCR2 receptor to activate downstream extracellular signal-regulated kinase 1/2 (ERK1/2) in neurons during neurodevelopment and neural repair 49,50 . Intriguingly, phosphorylated ERK1/2 (pERK1/2) has been paradoxically reported to both enhance HEVA71 replication 51 and initiate antiviral and anti-apoptotic responses; the latter in cooperation with Janus kinase 2 (JAK2) to phosphorylate STAT3 Fig. 1 Productive infection of primary neuronal cultures with HEVA71 causes neuronal cell death. a Immunostaining of UI-, S41-or C2-infected DIV7 neuronal cultures with antibodies against dsRNA and Tau. Scale bars represent 20 µm. Arrowheads indicated dsRNA-positive neurons. b Quantification of dsRNA-positive cells in infected DIV7 neuronal cultures at 6 hpi. Two hundred cells taken from random fields were manually counted for each condition. Percentages of dsRNA-positive cells were highest in S41-or C2-infected neurons but almost negligible in UI-infected controls. c Immunoblotting of cell lysates from S41-infected neuronal cultures for VP0 and VP2. β-actin was used as a loading control. VP0 and VP2 bands appeared at 6 hpi and continued to increase over the next 42 h. d Quantification of progeny viral particles release into supernatants of S41-or C2-infected neuronal cultures, where maximal viral titers were detected at 24 hpi in DIV7 neuronal cultures. Peak viral titers were similar in both infected neuronal cultures and RD. e Apoptotic percentages of caspase 3-or TUNEL-positive cells were highest in S41-or C2-infected neuronal cultures that were similar to the positive controls (staurosporine or TAC nuclease). Two hundred cells taken from random fields were manually counted for each condition. Scale bars represent 20 µm. MOI of 1 was used for all experiments. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3). Statistical analyses were performed with two-way ANOVA corrected using Bonferroni post-test. (****P < 0.0001). Source data are provided as a Source Data file.
at its serine 727 (pS727) and tyrosine 705 (pT705) residues 52,53 . The strong correlation of CXCL1 levels with disease severity is striking and indicated that the cytokine could promote viral infection.
To discriminate between the 2 contrasting roles, we exposed neurons to CXCL1 prior to infection. Pre-treatment of neurons with 25 units (U) or more of CXCL1 both accelerated and increased viral titers at 12 hpi in comparison to untreated controls (Fig. 4a). Expectedly, these were accompanied by an earlier drop in viral titer and a faster decline in neuronal viability ( Supplementary Fig. 11a). Production of viral proteins was similarly enhanced. Significantly higher levels of VP0 and VP2 were observed at 12 hpi instead of 24 hpi as seen in untreated controls (Fig. 4b, lanes 4, 7 and 10 versus lane 2; Fig. 4c).
An earlier and higher phosphorylation of ERK1/2 and STAT3(pS727) coincided with the accelerated viral replication in CXCL1 pre-treated neurons (Fig. 4b, lanes 4, 7 and 10 versus lane 1; Fig. 4d; Supplementary Fig. 11b, c). Furthermore, absence of JAK2 and STAT3(pT705) in both untreated and CXCL1pretreated neuronal cultures at time points after peak viral protein production argues against the involvement of pERK1/2 in initiating anti-viral responses via JAK2 (Fig. 4b, lanes 2-12; Supplementary Fig. 11d). Neuronal apoptosis coinciding with the timing of higher pERK1/2 levels also supports the role of the latter in enhancing viral replication (Fig. 1e).
Further supporting its involvement in enhancing viral replication only in neurons of neuronal cultures, CXCL1 pre-treatment did not accelerate viral replication, neuronal death, ERK1/2 activation and JAK2 degradation in infected astrocytes (Supplementary Fig. 14a-e), which do not express CXCR2 (Supplementary Fig. 14f). Surprisingly, viral synthesis and ERK1/2 activation increased in neuronal cultures at 48 hpi and was accompanied by strong reduction of JAK2 and STAT3(pT705) (Fig. 6b, lane 6; Fig. 6c, d; Supplementary Fig. 13c-f). This suggested that residual CXCL1 signaling through incompleted knockout of CXCR2 could eventually still allow viral replication to continue. Collectively, the results confirmed that astrocytic CXCL1 acts directly on neuronal CXCR2 to enhance viral replication in neurons.
HEVA71 VP4 elicits CXCL1 release and is required for viral replication in neurons. Next, we wondered which viral protein(s) could trigger CXCL1 release. EGFP-tagged viral protein (VP1 to VP4, 2A to 2C, 3A to 3D) or EGFP alone (control) were individually expressed lentivirally in neuronal cultures and astrocytes (Supplementary Fig. 15a, b). CXCL1 assays at 6 dpi indicated only VP4 was able to increase CXCL1 levels in both cultures ( Fig. 7a; Supplementary Fig. 15c). Incidentally, neurons expressing VP4 exhibited cell death whereas viral 3C protease (3C pro ) was responsible for astrocytic cell death ( Supplementary Fig. 15a, b). None of the other viral proteins tested showed similar effects.
Given the diminished ability of VP4-mut to stimulate CXCL1 release and cause neuronal cell death, we wondered if VP4-mut could inhibit infectivity of HEVA71 viruses carrying this mutation. We observed that ERK1/2 was not phosphorylated in astrocytes expressing either form of VP4 ( Supplementary  Fig. 15h). Instead, pERK1/2 could clearly be detected in neuronal cultures (Fig. 7c, d). Moreover, similar to their differences in eliciting CXCL1 release, VP4-mut was less potent at causing phosphorylation of ERK1/2 (and also downstream STAT3-pS727) as compared with wildtype VP4 (Fig. 7c, d; Supplementary  Fig. 15i, j). Only VP4 was able to induce elevation in ERK1/2 levels, which was not seen with VP4-mut or EGFP expression ( Supplementary Fig. 15i). Additionally, VP4-mut expression did not cleave caspase 3 whereas wildtype VP4 triggered robust caspase 3 cleavage ( Fig. 7c and e). Notably, expression levels of both VP4 variants were similar to those observed in infections using whole virions ( Supplementary Fig. 15k, l). This indicated that these findings were not due to over-expression artefacts. Moreover, only 3C pro (not either of the VP4 variants) triggered caspase 3 cleavage when singly expressed in astrocytes a Levels of the cytokines in cell culture supernatants from S41-or C2-infected neuronal cultures collected over a 96 h infection course. Uninfected neuronal culture was used as control. CXCL1 showed the highest increase during HEVA71 infection. G-CSF exhibited a delay in its upregulation in contrast to MCP-1, MIP1α, RANTES and CXCL1. Dotted lines represent the peak cytokine concentration that mostly corresponded to peak viral titers detected at 24 hpi for DIV7 neuronal cultures. b Immunostaining of S41-, C2-or UI-infected astrocytes with antibodies against dsRNA and GFAP. Scale bars represent 20 µm. Strong dsRNA-positive signals were detected in HEVA71-infected astrocytes (arrowheads) but not in UI controls. c Quantification of dsRNA-positive cells in infected astrocytes at 24 hpi. Two hundred cells taken from random fields were manually counted for each condition. d CXCL1 was elevated in cell culture supernatants of HEVA71-infected astrocyte cultures but not in UI controls at 24 hpi. MOI of 1 was used for all experiments. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3). Statistical analyses were performed with two-way ANOVA corrected using Bonferroni post-test. (****P < 0.0001). Source data are provided as a Source Data file.
To determine if VP4 mutations indeed attenuate HEVA71 infection in neurons, we generated VP4-mutant HEVA71 viruses using an infectious clone 59,60 . In human RD cells, both wildtype and VP4-mut viruses displayed similar viral fitness (viral replication, translation and peak viral titer) ( Supplementary  Fig. 16). VP4-mut viruses were significantly attenuated in infecting neuronal cultures in contrast to their wildtype counterparts. VP4-mut viruses displayed a long lag in viral protein synthesis and took a significantly longer time to reach peak viral titers similar to wildtype viruses (>144 versus 24 hpi) (Supplementary Fig. 17). Likewise, CXCL1 levels from VP4-mutant virus infected cultures took 6 times longer (144 versus 24 hpi) to reach similar levels elicited by wildtype viruses-something that was not observed in infected RD cells ( Supplementary Fig. 18). The lower extent of CXCL1 production commensurated with reduced viral protein synthesis and pERK1/2 levels, accompanied by detection of JAK2 and STAT3(pT705) in VP4-mutant HEVA71infected cultures in contrast to their wildtype counterparts ( Fig. 8a-c; Supplementary Fig. 19a, b). Of note, ERK1/2 levels were also lower in VP4-mutant HEVA71-infected cultures than in wildtype cultures, agreeing with our previous results using lentiviral VP4 expression ( Supplementary Fig. 19c).
Reverting to our mice model, we intracranially infected animals with wildtype, VP4-mut or UI viruses and monitored them afterwards for 15 days. Supporting our in vitro observations, mortality and the onset of clinical manifestations were significantly delayed in VP4-mut virus-infected mice (7 dpi) whereas they appeared earlier (3 and 4 dpi, respectively) in wildtype virusinfected mice (Fig. 9a, b). These changes were accompanied by significantly lower viral loads in the brains of VP4-mut versus wildtype infected mice at 7 dpi (Fig. 9c). Moreover, CSF levels of CXCL1 in VP4-mut infected mice remained undetectable until after 6 dpi (versus 4 in wildtype) and only reached equivalent highest levels around 13 dpi (versus around 7 dpi in wildtype) (Fig. 9d). Only slight differences in sera CXCL1 levels were observed in mice infected with either virus (Fig. 9e). Interestingly, the delayed onset of CXCL1 secretion, severe clinical manifestations and mortality in VP4-mut versus wildtype infected mice was not due to the reversion of VP4-mut to its wildtype counterpart virus (Supplementary Fig. 20). For both viruses, detection of CXCL1 preceded the onset of clinical symptoms and fatality. Notably, no change was observed in sera or CSF CXCL1 levels of UI controls throughout 15 dpi, which is in agreement with the negligible viral titers, lack of fatality and clinical symptoms ( Fig. 9a-e).
Suppressing CXCL1 signaling with AZD5069 prolongs survival and reduces clinical severity in HEVA71-infected mice. Given the critical involvement of CXCL1 in establishing neuronal infections, we wanted to evaluate whether pharmacological pertubation of its signaling would interfere with and, hence, alleviate HEVA71-induced neuropathogenesis. AZD5069, a drug currently used for cancer and respiratory illness in humans, is a potent, selective and reversible CXCR2 antagonist 61,62 . Indeed, application of AZD5069 to CXCL1-exposed neuronal cultures effectively reduced ERK1/2 activation with no observable toxicity (Supplementary Fig. 22a, b). Single application of AZD5069 to infected neuronal cultures reduced pERK1/2 levels and delayed viral protein synthesis by at least 12 h in contrast to vehicle-treated controls ( Supplementary Fig. 22c, lanes 1-2 versus lanes 4-5; Supplementary Fig. 22d, e). AZD5069 has a reported half-life of 16 h 61 . In agreement with this, viral protein synthesis and increased pERK1/2 levels resumed from 24 hpi, accompanied by the gradual disappearance of JAK2 and STAT3(pT705) (Supplementary Fig. 22c, lanes 5-6; Supplementary Fig. 22d-g). Together, these results showed that a single dose of AZD5069 is already sufficient to interrupt and delay the replication of HEVA71 in neurons.
To assess if AZD5069 can improve infection outcome in vivo, we intracranially administered AZD5069 (10 mg/kg) or vehicle control (PBS) once every 2 days up till 5 dpi in our HEVA71infected mouse model. Strikingly, mortality and the onset of clinical symptoms were siginficantly delayed in AZD5069-treated infected mice as compared to vehicle control (7 and 8 dpi versus 3 and 4 dpi, respectively) ( Fig. 10a, b). Concomitant with the milder infection in AZD5069-treated mice, much lower viral titers in the brain were detected between 4 to 7 dpi as opposed to those treated with vehicle control (Fig. 10c).
Levels of CXCL1 in the sera and CSF did not differ between the treated and untreated groups (Fig. 10d, e), indicating that AZD5069 had no effect on CXCL1 release. Yet, immunoblotting of brain lysates revealed that AZD5069-treated infected mice showed reduced levels of viral protein synthesis and phosphorylated pERK1/2 in comparison to vehicle-treated controls at 7 dpi (Fig. 10f, lane 2 versus lane 1; Fig. 10g, h). Likewise, caspase 3 cleavage was absent in AZD5069-treated infected mice, and JAK2 and STAT3(pT705) remained detectable (Fig. 10f, lane 2 versus lane 1; Fig. 10i-k). Of note, the detection of low percentages of CD45-and Ly6G-positive neutrophils in brains of both AZD5069-and PBS-treated infected mice at 7 dpi affirmed that reduced CNS damage upon AZD5069 treatment was not due to blockage of CXCL1-mediated neutrophil recruitment into CNS ( Supplementary Fig. 23). AZD5069 was not effective at curbing mortality, high clinical scores, viral loads and CXCL1-mediated signaling axis when given intravenously in S41-infected mice ( Supplementary Fig. 24), suggesting that AZD5069 was unable to cross the blood-brain barrier to enter the CNS and prevent CXCL1-mediated HEVA71 neuropathogenesis. Collectively, these results demonstrate that targeting CXCL1 signaling during severe HEVA71 infections is a viable strategy to limit neutrophilindependent CNS damage and improve infection outcome.

Discussion
Dysregulated cytokine production is strongly linked to severe HEVA71 infections with neurological complications 9 . However, the identity of CNS relevant cytokines and how they contribute to severe infections remain obscure [63][64][65] . We showed that productive HEVA71 infection of neurons and astrocytes elicits release of Fig. 3 CXCL1 is highly detected in CSFs of HEVA71 HFMD patients and mice with neurological deficits. a Survival rates and b mean clinical scores of mice infected with S41 or UI. S41-infected mice but not UI-infected mice displayed mortality and high clinical scores from 4 dpi. c High viral loads were detected in brains of S41-but not UI-infected mice (below LOD = 0.2) at 7 dpi. d, e CXCL1 levels were slightly elevated in d serum but highly elevated in e CSF of S41-infected mice but not in UI-infected mice that correlated with neuropathological deficits seen only in S41-infected mice. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3, 10 mice were used for each group per independent experiment). f CSF CXCL1 chemokine concentrations in severe HEVA71 HFMD patients significantly exceeds CXCL1 levels of other groups. Violin plots represent all analyzed values. Bold and dashed lines indicated median and the 25 th and 75 th percentile, respectively. Number of patients per group is indicated on the xaxis. Statistical analyses were performed with two-way ANOVA corrected using Bonferroni post-test. (*P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001). Exact P values for: moderate HEVA71 HFMD (CSF) = 0.0033, severe CA6 HFMD (CSF) = 0.0301, severe CA16 HFMD (CSF) = 0.0038 and JEV encephalitis (CSF) = 0.0401. Source data are provided as a Source Data file. Exact P values for VP2: 25 U/ml at 24 hpi = 0.001 and 100 U/ml at 24 hpi = 0.0215. d pERK1/2 to ERK1/2 ratios were higher in pretreated cultures versus controls in a CXCL1 dose-dependent manner at all timepoints. Exact P value for pERK1 to ERK1 ratio: 25 U/ml at 48 hpi = 0.0094. β-actin was used as a loading control for immunoblotting experiments. MOI of 1 was used for all experiments. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3). Statistical analyses were performed with two-way ANOVA corrected using Bonferroni post-test. (*P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001). Source data are provided as a Source Data file.
Perturbation of CXCL1 signaling using 2 distinct pharmacological strategies and, independently, via a genetic approach in vitro and in vivo effectively reduced viral replication while also preventing neuronal cell death and improving infection outcome. AZD5069 treatment of HEVA71-infected mice delayed mortality and the onset of clinical symptoms (Supplementary Fig. 26). Bar charts below show the quantifications for each protein. c VP0 and VP2 levels decreased at 12 and 24 hpi with ERK1/2 inhibition. d pERK1/2 to ERK1/2 ratios were reduced at all timepoints in FR180204-treated cultures. Exact P value for pERK1 to ERK1 ratio with FR180204 at 48 hpi = 0.0001. β-actin was used as a loading control for immunoblotting experiments. MOI of 1 was used for all experiments. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3). Statistical analyses were performed with two-way ANOVA corrected using Bonferroni posttest. (***P < 0.001 and ****P < 0.0001). Source data are provided as a Source Data file.
Collectively, these results uncover the involvement of CXCL1 in causing neurological complications and highlight the feasibility of pharmacologically interfering its signaling in the treatment of severe HFMD infections.
Whilst low CXCL1 levels in sera were previously detected in HFMD patients, its levels in CSF and relation to disease severity has not been hitherto appreciated 66 . Our study demonstrates the chemokine's involvement as a neuronal specific pro-viral factor to promote HEVA71's infection in neurons. Although CXCL1 has been reported to facilitate CNS infiltration of neutrophils during mouse hepatitis virus (JHMV) 41    the lack of contribution by CXCL1-recruited infiltrating neutrophils in causing HEVA71-induced CNS damage. Indeed, this is consistent with previous reports showcasing sparse neutrophil presence in CNS during HEVA71 infection [67][68][69] , even though some studies have reported otherwise 70,71 . Despite this controversy, our additional observations that neuronal cell death occurs in our cultures in the absence of neutrophils and that CXCL1 directly enhances viral replication in neurons, strongly indicate viral replication but not neutrophils as the primary cause of damage.
The structural protein VP4 is known to participate in viral entry 72,73 . However, the possibility of its involvement in directly targeting host functions during enteroviral infections has not been reported. We observe that VP4 expression in astrocytes is sufficient to trigger the release of large amounts of CXCL1 from these cells. Myristylation-deficient VP4, while still capable of eliciting lower amounts of CXCL1 release, attenuated HEVA71's ability to infect neurons. These findings are significant as they established a role for VP4 as a proviral factor important for establishing HEVA71 infections in the CNS. Understanding how VP4 regulates astrocytic CXCL1 expression and how it promotes HEVA71 replication in neurons but not other cells are topics requiring further investigation.
Expression of CXCR2 in neurons but not astrocytes and release of CXCL1 in the latter but not former indicates an unidirectional axis of signaling for CXCL1 from astrocytes to neurons during the course of virus infection. Indeed, perturbing this signaling disrupts HEVA71 replication in neurons but not astrocytes. CXCL1 and CXCR2 have been implicated in other forms of CNS injury and blockade of CXCR2 signaling has been shown to be neuroprotective 74 . In agreement with previous studies, activation of CXCR2 in neurons increased ERK1/2 phosphorylation 75 . Activated ERK1/2 is required for enhancing viral replication 51,76 . Supporting this, VP4-mut HEVA71 viruses that were less competent at eliciting CXCL1 release and ERK1/2 phosphorylation exhibited lower rates of replication. This is in agreement with the notion that ERK1/2 activation must exceed a certain threshold before HEVA71 can efficiently infect neurons 57,58 . Elevated ERK1/2 activation can also promote apoptosis. Indeed, previous studies observed positive correlation between pERK1/2 levels with neuronal death in neurodegenerative diseases 77 . The VP4-CXCL1-CXCR2-ERK1/2 axis appears to be the predominant if not sole mechanism for triggering neuronal but not astrocyte apoptosis. Supporting this, VP4-mut, which was significantly less able to activate ERK1/2, could not induce neuronal apoptosis. In comparison, viral 3C pro expression induces astrocyte but not neuronal apoptosis, possibly by activating caspase 8 as shown in other studies 30,78 . Thus, coordination between different viral and host proteins may be required to regulate apoptosis in different cell types during viral infection.
Apart from ERK1/2 activation, our study documented significant ERK1/2 upregulation during HEVA71 replication in neuronal cultures, particularly during VP4 presence. Reinforcing this notion, VP4-mut was found to maintain basal ERK1/2 levels in neuronal cultures. As previously reported, many viruses along with HEVA71, such as human immunodeficiency virus (HIV), hepatitis C virus (HCV) and Kaposi's sarcoma-associated herpesvirus (KSHV) can target ERK1/2 pathways as a means to manipulate cellular function and to regulate viral infection and replication, albeit it should be noted that ERK1/2 activation rather than its upregulation was documented in those studies [79][80][81][82] . Given that ERK1/2 elevation was not present in infected astrocytes, we can only suspect that myristoylated VP4 could cause ERK1/2 accumulation specifically in neurons, which would require further investigation.
The lack of a HEVA71 vaccine has prompted intense studies to identify effective antivirals against severe HEVA71 infections 4 . We showed that AZD5069, a reversible and selective CXCR2 antagonist, efficiently inhibited viral replication and significantly improved clinical severity and survival rates in HEVA71-infected mice. AZD5069 is safe and well-tolerated in humans and is currently in used for the treatment of several diseases [83][84][85] . However, our results demonstrating the poor permeability of intravenously administered AZD5069 along with its documented low bioavailability across blood-brain barrier following oral administration 86 limits its direct application against HEVA71 pathogenesis. Testing AZD5069 derivatives that can effectively cross the blood-brain barrier and antagonise CXCL1 signaling in the brain could be a potential therapeutic avenue for HEVA71associated neurological deficits. Study design. We aimed to establish a representative cell model of HEVA71 that closely mimics brain physiology so as to gain a deeper understanding of the mechanisms in HEVA71 neuropathogenesis. We established that cultured primary rat neurons and astrocytes closely recapitulate the key features of HEVA71 infection, thereby establishing its usefulness as an appropriate cell line for modeling HEVA71 neuropathogenesis. We then elucidated the cytokine profiles of HEVA71infected neuronal cultures so as to identify highly elevated cytokines. Highly elevated CXCL1 in vitro was also detected in CSF of in vivo models and severe HFMD patients; and was subjected to pre-treatment studies in HEVA71-infected neuronal cultures, in which subsequent downstream studies such as protein and viral quantification studies were carried out. We next investigated and affirmed the mechanistic action of CXCL1 in neuronal cultures via protein studies involving lentiviral CRISPR-Cas9 CXCR2 abrogation and ERK1/2 suppression with small molecules. Downstream lentiviral studies involving expression of individual HEVA71 structural and non-structural proteins were performed in neuronal cultures to discover relevant viral protein(s) that induce CXCL1 secretion and/or manipulate CXCR2 signaling cascade in order to trigger HEVA71 neuropathogenesis mechanisms. Further mutation studies were carried out in relevant viral protein(s) to understand how they regulate CXCR2 signaling in vitro and in vivo. All data was collected from 3 independent experiments consisting of triplicates. Primary neurons and astrocyte cell culture. Adult female rats (MPF HanTAc: WH, Wistar) with their corresponding litters (sex undetermined, 8 to 10 pups per time) were purchased from the Centre for Animal resources and housed in temperature controlled (23 ± 1°C), individually ventilated cages on a 12/12-hour light/ dark cycle (7 am to 7 pm) with access to food and water. Primary rat hippocampal neurons or astrocytes were isolated from postnatal day 0 rat pups. Dissected cortices or hippocampi were trypsinized for 30 min prior to addition of serum medium consisting of Modified Eagle's Medium (MEM; Sigma-Aldrich) supplemented with 5% FCS, 2 mM L-alanyl-L-glutamine (Millipore), 1 x MEM Vitamine (Sigma-Aldrich), Mito+Serum extender (Corning Costar) and 3.8 g/L D-glucose to stop digestion. Digested tissues were triturated with a fire-polished Pasteur pipette and passed through a 40 µm cell strainer (Corning Costar) to remove cell debris. Resulting solution was subjected to centrifugation (500 x g) for 5 min at 25°C to obtain a clean cell pellet that was resuspended in serum medium. Hippocampal neurons were plated directly on poly-D lysine (PDL; Sigma-Aldrich) coated coverslips in 12-well plates (35,000 cells/well) or PDL-coated 96-well plates (7,000 cells/well) or T-25 culture flasks (100,000 cells/flask) in neuronal plating DMEM/ F12 medium (Sigma-Aldrich) and 1 x B27 (Invitrogen). Medium was changed to fresh plating media supplemented with 1 x FUDR (Sigma-Aldrich) after 3 days. Astrocytes were plated on PDL-coated coverslips in 12-well plates (35,000 cells/ well) or PDL-coated 96-well plates (7,000 cells/well) in astrocyte plating media (DMEM with 10% FCS).

Methods
Animal studies. All animal work was carried out in accordance with the IACUCapproved protocol (R20-14766) and in Animal Biosafety Level 2 (ABSL2) facility. Adult female mice (BALB/c/CAnNTac) with their corresponding litters (sex undetermined) were purchased from the Centre for Animal resources and housed in temperature controlled (23 ± 1°C), individually ventilated cages on a 12/12-hour light/dark cycle (7 am to 7 pm) with access to food and water. 1-day-old BALB/c suckling mice (sex undetermined; 26 per group) (Invivos) were infected intracranially with HEVA71 S41 strain or UI or VP4-mutant HEVA71 (2 ×10 7 PFU per mouse) and survival rates and clinical manifestations were monitored for 15 days post infection (dpi) as previously described 38 . For AZD5069 studies, AZD5069 (10 mg/kg) or phosphate-buffered saline (PBS) was administered intracranially at 0, 2 and 4 dpi or intravenously at 0, 2, 4 and 6 dpi following UI or S41 infection in mice (sex undetermined; 25 per group) and monitored for 15 days post infection for survival rates and clinical manifestations. For intracranial Ly6G neutralization studies, IgG isotype (TNP6A7) or anti-Ly6G antibody (BP0075-1) was purchased from BioXcell and administered intracranially at a dose of 1 mg/kg daily from 0 to 5 dpi after UI or S41 infection in mice (sex undetermined; 25 per group). These mice were then monitored for their clinical scores and survival rates till 8 dpi. Severity of clinical symptoms in infected mice were scored by a mice clinical Fig. 9 Mutating HEVA71 VP4 reduces CXCL1 signaling and associated HEVA71 replication and neuropathogenesis in mice. a-e Infection of mice with wildtype or VP4-mutant HEVA71 virons. a Survival rates and b mean clinical scores of mice infected with wildtype, VP4-mutant or UI virus. Wildtype and VP4-mutant HEVA71-infected mice displayed mortality and high clinical scores from 4 and 7 dpi, respectively. UI-infected mice were healthy till 15 dpi. c VP4-mutant HEVA71-infected mice had lower viral loads detected in their brains at 7 dpi. UI-infected mice were used as controls. Negligible viral loads were detected below limit of detection (LOD = 0.2). d, e CXCL1 levels were highly elevated in d CSF but only slightly elevated in e serum of both wildtype and VP4-mutant HEVA71-infected mice but not in UI-infected mice. Wildtype HEVA71-infected mice showed increasing CXCL1 from 4 dpi but those for VP4-mutant-infected mice emerged at 7 dpi. f Brain lysates of wildtype, VP4-mutant or UI-infected mice were analyzed by immunoblotting at 7 dpi. Infection with VP4-mutant HEVA71 delays the expression of HEVA71 viral proteins. Immunoblot representative of 3 biological repeats is shown. Bar charts below show quantifications for each protein. g VP0 and VP2 levels decreased at 7 dpi with VP4-mutant HEVA71. Exact P values for: VP0 = 0.0009 and VP2 = 0.0002. h pERK1/2 were reduced in VP4-mutant HEVA71-infected mice. Exact P values for: pERK1 = 0.0009 and pERK2 = 0.0012. i Cleaved caspase 3 was detected in wildtype HEVA71-infected mice but not VP4-mutant HEVA71-infected mice at 7 dpi. Exact P values for: caspase 3 = 0.0254 and cleaved caspase 3 = 0.0008. β-actin was used as a loading control for immunoblotting experiments. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3). Statistical analyses were performed with two-way ANOVA corrected using Bonferroni posttest. (**P < 0.01, ***P < 0.001 and ****P < 0.0001). Source data are provided as a Source Data file.
scoring system (Supplementary Table 3). Cerebrospinal fluids or sera were collected daily from 0 to 7 or 15 dpi from surviving mice of all groups and quantified for levels of CXCL1 (sex undetermined; 10 per group). Mice brains were harvested from 0 to 15 dpi for subsequent quantification of viral load via plaque reduction assays or for immunohistochemistry and immunoassays.
Immunohistochemistry. To examine neuropathology and expression of viral antigens and CXCL1, brains from both groups were harvested at 3 to 6 dpi (if mice died at early timepoints due to S41 infection) and 7 dpi (if alive) and fixed with 4% paraformaldehyde (PFA) for 1 week at 4°C. Sectioning of 80 µm thick free floating coronal sections through the prefrontal cortex and hippocampus were performed using a vibratome (VR1200S, Leica). Sections were collected in cold PBS. Three consecutive sections through each region of interest were alternately stained using Nissl staining or with respective antigen-specific antibodies.
Nissl staining was used to evaluate neuronal cell death. Respective sections were mounted on gelatin-coated slides and dried at room temperature. Following treatments with 0.5% crystal violet solution (Sigma-Aldrich), sections were rehydrated in increasing serial concentrations of ethanol and xylene. As a final step, coverslips were applied to the sections. Sections were then digitized at a resolution of 1280 ×960 pixels under a Nikon Eclipse Ci microscope equipped with DS-F12 camera using the NIS Elements software version 5.21.00.
To examine the presence of viral antigens, respective sections were treated with 0.3% hydrogen peroxide (Sigma-Aldrich) to remove endogenous catalase activity. After blocking in 3% bovine serum albumin (BSA, Sigma-Aldrich), sections were incubated with mouse anti-HEVA71 antibody (1:200, Millipore, MAB979, clone 422-8D-4C-4D) prior to another overnight incubation with biotinylated antimouse secondary antibody (1:400, Santa Cruz, sc2031) at 4°C. Antibodies used were diluted in BSA. Immunolabelling was developed using the horse radish peroxidase (HRP) conjugated avidin-biotin kit (Vectastain, Vector laboratories) with 0.1% diaminobenzidine as a chromogen. Sections were dried at room temperature on gelatin-coated slides and treated with increasing doses of ethanol and xylene before coverslips were applied. Sections were visualized under a Nikon Eclipse Ci microscope attached with DS-F12 camera through the NIS Elements software.
To examine the expression of CXCL1 in neurons, astrocytes and microglia, respective sections were similarly blocked in 3% BSA and incubated with respective combinations of rabbit CXCL1 antibody  (Supplementary Fig. 10). Sex, number, age of participants and suspected etiologies per group is summarized in Supplementary Table 4. Clinically-defined HFMD and other viral cases were confirmed using one of the laboratory tests, including viral genotyping and HEVA71, CA6, CA16 or JEV IgM antibody detection. Informed consent was obtained from the parents/guardians of all the recruited subjects from KK Women's and Children's hospital to use the sera samples collected. For CSF samples, only verbal consent was obtained. No compensation for participation was provided. Resultant CSF and sera samples were subjected to CXCL1 cytokine measurements.
Virus infection studies. Neuronal cultures, astrocytes and RD cells were infected with either HEVA71 S41 or HEVA71 C2 strain at MOI of 1, 10 and 50. CA6 and CA16 were used at MOI of 1 to infect neuronal cultures. After incubation with the viruses for an hour, the cell monolayers were washed with phosphate buffered saline (PBS) twice prior to further incubations in virus propagation media. At various time points post-infection, respective cells or supernatants were harvested and processed for cell viability assays, apoptosis assays, plaque assays, immunoblot analyses, quantitative real-time PCR assays, immunofluorescence assays and cytokine analyses. HEVA71 VP4 mutagenesis. A glycine (GGC) to alanine (GCC) mutation was carried out at the 5 th position of VP4 gene region in the HEVA71 infectious clone. Part of the HEVA71 infectious clone excluding VP4 genes was amplified by PCR with the appropriate primers using Q5 ® High-Fidelity master mix protocol. Primers used were purchased from Integrated DNA Technologies (IDT): HEVA71 forward (5'-TCTCCATCCGCTGAG-3') and reverse (5'-GTTTGATTGTATT-GAGGGTTAAT-3'). Subsequently, VP4 gene of HEVA71 infectious clone was mutated and amplified with pair of VP4-mut primers (IDT): VP4-mut forward (5'-TCAATACAATCAAACATGGCCTCACAGGTGTCT-3') and reverse (5'-CTCAGCGGATGGAGACTTCAGTGGTGCAGCCAT-3'). Of note, the forward primer contained the intended mutation. After agarose gel extraction and purification, both fragments were fused with the In-Fusion ® HD cloning kit according to manufacturer recommendations. Generated VP4-mut HEVA71 infectious clone was verified via automated sequencing by Bio Basic Asia Pacific Ltd (Singapore) before upscaling and usage to produce infectious mutant viral particles. Apoptosis assays. Following various time points (24 or 48 hpi) after HEVA71 infections (S41, C2 or UI) in DIV7 neuronal cultures, respective cells were fixed with 3.7% paraformaldehyde (PFA) prior to TUNEL assays or immunostainings with caspase 3 antibody. For TUNEL assays, the TiterTACs TM Colorimetric Apoptosis Detection Kit (Trevigen) was utilized according to manufacturer's instructions. Corresponding fluorescence was measured at absorbance wavelength of 450 nm. TAC nuclease and uninfected cells were utilized as a positive and negative control, respectively. For caspase 3 immunostainings, staurosporine (1 uM) and uninfected cells were used as corresponding positive and negative controls.
Viral plaque reduction assays. A day prior to performing plaque assays, RD cells were seeded in 24-well plates (50,000 to 100,000 cells/well). Cell supernatants or homogenized animal brain tissues from various infection studies were serially diluted 10-fold, ranging from 10 −1 to 10 −7 in DMEM with 2% FCS. Respective diluted solutions were incubated with RD cells from 1 to 1.5 h to allow for viral adsorption. Plates were then washed two times with PBS and added with overlay media of DMEM with 2% FCS and 1% carboxymethylcellulose (CMC). After a 4-day incubation period, cells were fixed and stained with 10% PFA supplemented Fig. 10 Suppressing CXCL1 signaling by AZD5069 treatment improved survival and clinical scores that correlated with reduced HEVA71 viral load in infected mice. a Survival rates and b mean clinical scores of mice infected with S41 or UI with or without AZD5069 treatment, respectively. Arrowheads indicate time points of AZD5069 administration. S41-infected mice displayed mortality and high clinical scores from 4 dpi. AZD5069 administration to S41infected mice delayed mortality and high clinical scores until 8 dpi. UI-infected mice were healthy till 15 dpi. c AZD5069 treatment lowered viral loads in brains of S41-infected mice from 4 to 7 dpi. d, e CXCL1 levels were slightly elevated in d serum but highly elevated in e CSF of S41-infected mice with or without AZD5069 but not in UI-infected mice. f Brain lysates of S41-or UI-infected mice with or without AZD5069 treatment were subjected to immunoblotting analyses at 7 dpi. AZD5069 treatment delays the expression of HEVA71 viral proteins. Immunoblot representative of 3 biological repeats is shown. Bar charts below show quantifications for each protein. g VP0 and VP2 levels decreased at 7 dpi with AZD5069 treatment. Exact P values for: VP0 = 0.0006 and VP2 = 0.0009. h pERK1/2 and ERK1/2 levels were reduced in AZD5069-treated mice. Exact P values for: ERK1 = 0.0002, ERK2 = 0.0013, pERK1 = 0.0001 and pERK2 = 0.0007. i JAK2 levels persisted in CXCR2-suppressed mice with S41 infection. Exact P values for: JAK2 = 0.001. j AZD5069 treatment increased phosphorylation of STAT3 at pT705 but decreased pS727 levels at 7 dpi. Exact P values for: pT705 = 0.0007. k Cleaved caspase 3 was detected only in S41infected mice without AZD5069 treatment at 7 dpi. Exact P values for: caspase 3 = 0.0013 and cleaved caspase 3 = 0.0015. β-actin was used as a loading control for immunoblotting experiments. Data are expressed as the average with error bars representing ±SD from 3 independent experiments (n = 3). Statistical analyses were performed with two-way ANOVA corrected using Bonferroni post-test. (**P < 0.01, ***P < 0.001 and ****P < 0.0001). Source data are provided as a Source Data file.
Cell extract preparations and immunoblot analyses. Cell extracts were prepared with 4 x NuPAGE LDS Sample Buffer (Thermo Fisher Scientific). Quantification of respective cell lysates were performed via Pierce Protein Assay Reagent (Thermo Fisher Scientific) at a wavelength of 660 nm. 10 µg of protein samples were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad) via the Trans-Blot Turbo Transfer System (Bio-Rad). Membranes were blocked with blocking buffer consisting of 5% skim milk (FairPrice) in TBST (150 mM Tris-HCl, pH 7.4, 1.5 M NaCl, 0.5% Tween 20) for 30 min at 25°C. Membranes were incubated with primary antibodies diluted with blocking buffer overnight at 4°C. Following three 5-minute TBST washes, membranes were incubated with secondary antibodies. Protein bands were visualized with an enhanced chemiluminescent substrate (Thermo Fisher Scientific) using the C-DiGit Chemiluminescence Western Blot Scanner (LI-COR). Densitometric quantification was performed using ImageJ v1.48 freeware. Band intensities were normalized to actin.
Immunoflourescence assays. Cells were fixed with 3.7% PFA, and permeablised using 0.3% Triton-X100 (Sigma-Aldrich). After blocking with 10% normal goat serum (NGS; Millipore), cells were incubated with their respective primary antibody combinations for 1 h at 25°C. After three 10-minute washes in PBS, coverslips were incubated with secondary antibodies for another hour at 25°C and washed again thrice with three PBS washes with 10-minute intervals. As a final step, coverslips were mounted with Fluoro-Gel II with DAPI (Electron Microscopy Sciences) on glass slides and visualized using a Zeiss Axio Observer Z1 Microscope equipped with a motorised stage. Images were analyzed on Zeiss Zen 3.4.
The following primary antibodies were utilized for immunostainings: mouse anti-dsRNA antibody Cytokine analyses. Culture supernatants from HEVA71-infected neurons, HEVA71-infected astrocytes and lentiviral-infected neurons or CSF and sera taken from mice or human clinical cases over various time/day points were assayed using the respective Bio-Plex Pro rat cytokine 23-plex kit (Bio-Rad) or rat CXCL1 ELISA kit (Aviva Systems Biology) or human CXCL1 ELISA kit (ThermoFisher Scientific) according to manufacturer's instructions. Data from the Bio-Plex assays were processed using Bio-Plex Manager software version 6.1 (Bio-Rad). Cytokines with outof-range readings that were below its detection limits were recorded as 0.1 pg/ml. Compound preparations. Recombinant rat CXCL1 (25 µg) was purchased from Peprotech and diluted in water to yield 100,000 U/ml stock solution. FR180204 (5-(2-phenylpyrazolo[1,5-a]pyridin-3-yl)−1H-pyrazolo [3,4-c]pyridazin-3-amine) was obtained from Sigma-Aldrich and diluted in water to obtain a 10 mM stock solution. AZD5069 (10 mg) was purchased from MedChem Express and diluted in PBS to achieve a 25 mM stock solution. Stock solutions were stored at −20°C.
Compound treatment assays. Cultures were pretreated with a range of CXCL1 concentrations (0 to 100 U/ml) for 24 h prior to HEVA71 S41 infection at MOI of 1. Supernatants of DIV7 neuronal cultures were collected at 12, 24 and 48 hpi and plaque reduction assays were performed for viral titer quantifications. Lysates from DIV7 neuronal cultures were also harvested and subjected to immunoblot analyses. DIV7 neuronal cells were separately seeded in 96-well plates and treated with a range of doses of FR180204 (1.25 to 20 µM) or AZD5069 (0 to 750 nM) for 48 h, before relevant cell viabilities for each dosage were measured via cell viability assays. Following 100 U/ml of CXCL1 pre-treatments of DIV7 neurons, cells were infected with HEVA71 (S41) for an hour. After washing twice with PBS, 5 µM of FR180204 or 750 nM of AZD5069 was added in fresh neuronal culture maintenance media to HEVA71-infected neurons, where cells were further incubated for 12, 24 and 48 h prior to cell lysis and subsequent immunoblotting assays.
Lentiviral transduction of HEVA71 proteins. Individual HEVA71 proteins were separately expressed in DIV7 neuronal cultures. Specifically, cDNAs encoding fulllength HEVA71 (S41) VP1, VP2, VP3, VP4, VP4-mut, 2 A, 2B, 2 C, 3 A, 3B, 3 C or 3D were separately cloned into the FHUG-EGFP backbone (Addgene, plasmid #14883, a kind gift from Dr. David Baltimore) using the EcoRI and AgeI sites. Primers (IDT) used are indicated in Supplementary Table 5. The P2A cleavage sequence was inserted between the viral proteins and EGFP. Following their expression, the EGFP moiety is cleaved to allow for the expression of native viral proteins. VP4-mut cDNA was obtained via a site-directed mutagenesis (MGXXXS to MAXXXS) with In-Fusion Cloning kit (Takara) according to manufacturer's instructions (Supplementary Table 5). Empty FHUG-EGFP plasmid was utilized as EGFP expression control. Production of individual lentiviruses with each HEVA71 protein were performed as mentioned above and stored in 100 µl aliquots at −80°C until use. Neuronal cultures were fixed for immunofluorescence staining or lysed for immunoblotting at 6 days after lentiviral infection. Culture supernatants were also collected and subjected to CXCL1 cytokine analyses.
Immunoprecipitation assays. Using In-Fusion Cloning kit (Takara), a V5 tag was introduced from pcDNA3.1/nV5-DEST vector (Addgene, plasmid #12290010) into FHUG-EGFP-VP4 and FHUG-EGFP-VP4-mut according to manufacturer's instructions. 200 µl of concentrated lentiviruses expressing VP4 or VP4-mut tagged with V5 were then used to infect DIV1 neurons seeded in T-25 flasks. Empty FHUG-V5 plasmid was used as a control. After 6 days post infection, cells were washed with PBS twice and lysed with 500 µl of HEPES Lysis Buffer TBST (50 mM HEPES, 150 mM NaCl, 1 mM EDTA, 1% Triton-X100, pH 7.2) supplemented with protease inhibitor cocktail (100X; Thermo Fisher Scientific). Lysates were clarified by centrifugation at 10,000 g for 10 min at 4°C. Resultant supernatants were collected and added with 1 µl of anti-V5 antibody (1:200, Abcam, ab95038) for subsequent incubations at 4°C for 3 to 4 h. 30 µl of agarose beads (Thermo Fisher Scientific) were then washed three to five times with HEPES Lysis Buffer and added to each antibody-lysate mixture for another 1 to 2 h of incubation. Following this, resultant mixtures were spun down at 1000 x g at 1 min at 4°C and supernatants were removed. Beads were washed thrice in HEPES lysis buffer through inverting them 6 times and removing lysis buffer via centrifugation at 12,000 x g for 1 min at 4°C. Immune complexes were then eluted from beads by adding 2 x NuPAGE LDS Sample Buffer (Thermo Fisher Scientific) and resolved by subsequent SDSpolyacrylamide gel electrophoresis. Proteins were transferred to nitrocellulose membranes (Bio-Rad) and further incubated with appropriate antibodies.
Fluorescence-activated cell sorting (FACS) analysis. Leukocyte and neutrophil populations for FACS analyses were obtained as described previously 88,89 . Brain and muscle tissues from individual animals were separately minced gently in RPMI 1640 medium (Thermo Fisher Scientific) with a razor blade and incubated with collagenase D (Roche; 1 mg/ml) and DNAse I (Roche; 50 µg/ml) for 30 min at 37°C. Digested tissues were passed through a 70 µm cell strainer (Corning Costar) to remove cell debris and to obtain single cell suspensions. Cells were spun down at 500 x g for 5 min at 4°C. The resultant cell pellets from brain tissues were resuspended in ice-cold 90% Percoll (Sigma Aldrich) and overlaid with 60% Percoll, followed by 40% Percoll and subsequently HBSS (Thermo Fisher Scientific) gently. Cell pellets obtained from muscles were resuspended with 70% Percoll and gently covered with 37% Percoll followed by 30% Percoll. All Percoll solutions were diluted in HBSS. Cells were isolated by centrifugation at 600 x g for 30 min at 4°C. Braking was not used for deceleration. The cell fraction (containing leukocytes) located between the 60% and 40% interphase (for brains) and 70% and 37% phases (for muscles) were carefully aspirated. HBSS was added to the cell suspensions, which were spun again at 550 x g for 10 min at 4°C. Resultant cell pellets were resuspended in 100 µl FACS buffer (Sigma Aldrich). Samples were stained with fluorescent isothiocyanate (FITC)-conjugated anti-mouse Ly6G antibody (1:100, BD Biosciences, BD551460), PE-Cy7-conjugated anti-mouse CD45 antibody (1:200, eBioscience, 25-0453-82) and Calcein violet AM (1:200, Invitrogen, C34858) in the dark at 4°C for 30 min. Cells were then pelleted by centrifugation at 400 x g at 4°C for 5 min and washed twice with FACS buffer prior to analyses using the FACSCalibur flow cytometer (BD Biosciences). Data collected were analyzed using FlowJo TM version 10.8.
Fluorescence-activated cell sorting (FACS) gating strategy. Cells of interest were gated on forward scatter area (FSC-A)/side scatter area (SSC-A) plot to remove cell debris. Single cells were subsequently gated on forward scatter height (FSC-H)/forward scatter area (FSC-A) and forward scatter width (FSC-W)/forward scatter area (FSC-A) density plots to exclude doublets. Live leukocytes were identified by selecting cells that are dually positive for Calcein violet AM and CD45. These cells were then further gated for neutrophils that are both CD45-and LY6Gpositive. Data were analyzed using FlowJo TM software and population frequencies were expressed as percent of single cell population.
Statistical analysis. Statistics were performed via GraphPad Prism 9 version 9.3.0. All experiments were performed three independent times with 3 technical replicates. Data are represented as means and standard deviations (SDs), calculated by Microsoft Excel 2016/version 16.56. Two-way ANOVA followed by post hoc test for multiple comparisons was utilized in determining statistical significance, as described in figure legends. p ≤ 0.05 was considered significant. N values are also indicated within figure legends and refer to biological replicates, whereas technical replicates (included in Source Data file) refer to repeated analysis of the same samples. Derived statistics correspond to analysis of averaged values across biological replicates, and not pooled technically and biological replicates.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The main data supporting the findings of this study are available within the article and its Supplementary Figures. The source data and P values underlying the figures are provided as a Source Data file. Additional details on datasets that support the findings of this study will be made available by the corresponding or first author upon reasonable request. Source data are provided with this paper.